diff --git a/blog/src/assets/blog-press.css b/blog/src/assets/blog-press.css
index 06f9c4d6b..0784e1105 100644
--- a/blog/src/assets/blog-press.css
+++ b/blog/src/assets/blog-press.css
@@ -45,7 +45,7 @@ body {
}
.wrap { max-width: 1120px; margin: 0 auto; padding: 0 24px; }
a { color: var(--ink); }
-img { max-width: 100%; height: auto; }
+img, video { max-width: 100%; height: auto; }
.skip-link { position: absolute; left: -9999px; top: 0; background: var(--ink); color: #fff; padding: 10px 16px; z-index: 99; }
.skip-link:focus { left: 0; }
@@ -182,11 +182,14 @@ a.tag:hover { background: var(--ink); color: var(--paper); }
.ph-art { text-align: center; display: grid; gap: 8px; }
.ph-art .ph-glyph { font-size: 46px; color: var(--c, var(--accent)); }
.ph-art .ph-note { font-family: var(--grot); font-weight: 700; font-size: 12px; letter-spacing: 0.24em; text-transform: uppercase; color: var(--ink-soft); }
+.lead-loop { margin: 1.4rem 0 .6rem; border: 1px solid #D9CFE0; border-radius: 16px; overflow: hidden; background: #FFF8E7; }
+.lead-loop video, .lead-loop img { display: block; width: 100%; height: auto; margin: 0; border: 0; }
+.lead-loop figcaption { font: 500 13px/1.5 "Space Grotesk", system-ui, sans-serif; color: #6B5878; padding: 10px 16px 14px; margin: 0; border-top: 1px solid #D9CFE0; background: #FCFAF0; }
.fig-inline { margin: 42px auto; max-width: 620px; }
-.fig-inline img { display: block; width: 100%; border: 3px solid var(--rule); border-radius: 6px; background: #fff; }
+.fig-inline img, .fig-inline video { display: block; width: 100%; border: 3px solid var(--rule); border-radius: 6px; background: #fff; }
.fig-inline figcaption { text-align: center; }
-.fig-inline:nth-of-type(odd) img { transform: rotate(-0.4deg); }
-.fig-inline:nth-of-type(even) img { transform: rotate(0.35deg); }
+.fig-inline:nth-of-type(odd) img, .fig-inline:nth-of-type(odd) video { transform: rotate(-0.4deg); }
+.fig-inline:nth-of-type(even) img, .fig-inline:nth-of-type(even) video { transform: rotate(0.35deg); }
.fig-inline.fig-placeholder { aspect-ratio: 15/7; }
/* post body layout */
diff --git a/blog/src/assets/media/chatgpt-dots-alternatives/lead-still.png b/blog/src/assets/media/chatgpt-dots-alternatives/lead-still.png
new file mode 100644
index 000000000..8646a24d7
Binary files /dev/null and b/blog/src/assets/media/chatgpt-dots-alternatives/lead-still.png differ
diff --git a/blog/src/assets/media/chatgpt-dots-alternatives/lead.gif b/blog/src/assets/media/chatgpt-dots-alternatives/lead.gif
new file mode 100644
index 000000000..2a3dc825e
Binary files /dev/null and b/blog/src/assets/media/chatgpt-dots-alternatives/lead.gif differ
diff --git a/blog/src/assets/media/chatgpt-dots-alternatives/lead.mp4 b/blog/src/assets/media/chatgpt-dots-alternatives/lead.mp4
new file mode 100644
index 000000000..2bc74155f
Binary files /dev/null and b/blog/src/assets/media/chatgpt-dots-alternatives/lead.mp4 differ
diff --git a/blog/src/assets/media/chatgpt-dots-alternatives/note-1-still.png b/blog/src/assets/media/chatgpt-dots-alternatives/note-1-still.png
new file mode 100644
index 000000000..39cc53ee7
Binary files /dev/null and b/blog/src/assets/media/chatgpt-dots-alternatives/note-1-still.png differ
diff --git a/blog/src/assets/media/chatgpt-dots-alternatives/note-1.gif b/blog/src/assets/media/chatgpt-dots-alternatives/note-1.gif
new file mode 100644
index 000000000..9af7ea6c1
Binary files /dev/null and b/blog/src/assets/media/chatgpt-dots-alternatives/note-1.gif differ
diff --git a/blog/src/assets/media/chatgpt-dots-alternatives/note-1.mp4 b/blog/src/assets/media/chatgpt-dots-alternatives/note-1.mp4
new file mode 100644
index 000000000..c3d6e7741
Binary files /dev/null and b/blog/src/assets/media/chatgpt-dots-alternatives/note-1.mp4 differ
diff --git a/blog/src/assets/media/chatgpt-dots-alternatives/note-2-still.png b/blog/src/assets/media/chatgpt-dots-alternatives/note-2-still.png
new file mode 100644
index 000000000..733105698
Binary files /dev/null and b/blog/src/assets/media/chatgpt-dots-alternatives/note-2-still.png differ
diff --git a/blog/src/assets/media/chatgpt-dots-alternatives/note-2.gif b/blog/src/assets/media/chatgpt-dots-alternatives/note-2.gif
new file mode 100644
index 000000000..872ec568b
Binary files /dev/null and b/blog/src/assets/media/chatgpt-dots-alternatives/note-2.gif differ
diff --git a/blog/src/assets/media/chatgpt-dots-alternatives/note-2.mp4 b/blog/src/assets/media/chatgpt-dots-alternatives/note-2.mp4
new file mode 100644
index 000000000..76b5f02ec
Binary files /dev/null and b/blog/src/assets/media/chatgpt-dots-alternatives/note-2.mp4 differ
diff --git a/blog/src/assets/media/grok-bot-pricing/lead-still.png b/blog/src/assets/media/grok-bot-pricing/lead-still.png
index 7cba402b0..355cf99bf 100644
Binary files a/blog/src/assets/media/grok-bot-pricing/lead-still.png and b/blog/src/assets/media/grok-bot-pricing/lead-still.png differ
diff --git a/blog/src/assets/media/grok-bot-pricing/lead.gif b/blog/src/assets/media/grok-bot-pricing/lead.gif
index b2b97572d..87a7e08b8 100644
Binary files a/blog/src/assets/media/grok-bot-pricing/lead.gif and b/blog/src/assets/media/grok-bot-pricing/lead.gif differ
diff --git a/blog/src/assets/media/grok-bot-pricing/lead.mp4 b/blog/src/assets/media/grok-bot-pricing/lead.mp4
index 7d92a91eb..22c87223e 100644
Binary files a/blog/src/assets/media/grok-bot-pricing/lead.mp4 and b/blog/src/assets/media/grok-bot-pricing/lead.mp4 differ
diff --git a/blog/src/assets/media/meta-muse-alternatives/lead-still.png b/blog/src/assets/media/meta-muse-alternatives/lead-still.png
index 98022304c..6c7cab064 100644
Binary files a/blog/src/assets/media/meta-muse-alternatives/lead-still.png and b/blog/src/assets/media/meta-muse-alternatives/lead-still.png differ
diff --git a/blog/src/assets/media/meta-muse-alternatives/lead.gif b/blog/src/assets/media/meta-muse-alternatives/lead.gif
index f8e0eb46e..d6b91b3ee 100644
Binary files a/blog/src/assets/media/meta-muse-alternatives/lead.gif and b/blog/src/assets/media/meta-muse-alternatives/lead.gif differ
diff --git a/blog/src/assets/media/meta-muse-alternatives/lead.mp4 b/blog/src/assets/media/meta-muse-alternatives/lead.mp4
index 610d3f107..2c1ed5d7e 100644
Binary files a/blog/src/assets/media/meta-muse-alternatives/lead.mp4 and b/blog/src/assets/media/meta-muse-alternatives/lead.mp4 differ
diff --git a/blog/src/assets/media/meta-muse-grok-bot-privacy/lead-still.png b/blog/src/assets/media/meta-muse-grok-bot-privacy/lead-still.png
index 430db710b..113dfe664 100644
Binary files a/blog/src/assets/media/meta-muse-grok-bot-privacy/lead-still.png and b/blog/src/assets/media/meta-muse-grok-bot-privacy/lead-still.png differ
diff --git a/blog/src/assets/media/meta-muse-grok-bot-privacy/lead.gif b/blog/src/assets/media/meta-muse-grok-bot-privacy/lead.gif
index bb51f8908..4a74447bb 100644
Binary files a/blog/src/assets/media/meta-muse-grok-bot-privacy/lead.gif and b/blog/src/assets/media/meta-muse-grok-bot-privacy/lead.gif differ
diff --git a/blog/src/assets/media/meta-muse-grok-bot-privacy/lead.mp4 b/blog/src/assets/media/meta-muse-grok-bot-privacy/lead.mp4
index 2624c1000..079a5394b 100644
Binary files a/blog/src/assets/media/meta-muse-grok-bot-privacy/lead.mp4 and b/blog/src/assets/media/meta-muse-grok-bot-privacy/lead.mp4 differ
diff --git a/blog/src/assets/media/meta-muse-vs-grok-bot/lead-still.png b/blog/src/assets/media/meta-muse-vs-grok-bot/lead-still.png
index 7131c3110..25f4591c0 100644
Binary files a/blog/src/assets/media/meta-muse-vs-grok-bot/lead-still.png and b/blog/src/assets/media/meta-muse-vs-grok-bot/lead-still.png differ
diff --git a/blog/src/assets/media/meta-muse-vs-grok-bot/lead.gif b/blog/src/assets/media/meta-muse-vs-grok-bot/lead.gif
index e6a718e46..65d344786 100644
Binary files a/blog/src/assets/media/meta-muse-vs-grok-bot/lead.gif and b/blog/src/assets/media/meta-muse-vs-grok-bot/lead.gif differ
diff --git a/blog/src/assets/media/meta-muse-vs-grok-bot/lead.mp4 b/blog/src/assets/media/meta-muse-vs-grok-bot/lead.mp4
index c81268299..8d3cf32e4 100644
Binary files a/blog/src/assets/media/meta-muse-vs-grok-bot/lead.mp4 and b/blog/src/assets/media/meta-muse-vs-grok-bot/lead.mp4 differ
diff --git a/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/lead-still.png b/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/lead-still.png
new file mode 100644
index 000000000..e1bac3340
Binary files /dev/null and b/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/lead-still.png differ
diff --git a/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/lead.gif b/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/lead.gif
new file mode 100644
index 000000000..012f057d6
Binary files /dev/null and b/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/lead.gif differ
diff --git a/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/lead.mp4 b/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/lead.mp4
new file mode 100644
index 000000000..a9963bdbb
Binary files /dev/null and b/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/lead.mp4 differ
diff --git a/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/note-1-still.png b/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/note-1-still.png
new file mode 100644
index 000000000..91ea7009e
Binary files /dev/null and b/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/note-1-still.png differ
diff --git a/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/note-1.gif b/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/note-1.gif
new file mode 100644
index 000000000..f06a1564f
Binary files /dev/null and b/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/note-1.gif differ
diff --git a/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/note-1.mp4 b/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/note-1.mp4
new file mode 100644
index 000000000..9f63badad
Binary files /dev/null and b/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/note-1.mp4 differ
diff --git a/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/note-2-still.png b/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/note-2-still.png
new file mode 100644
index 000000000..5f4c8b00e
Binary files /dev/null and b/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/note-2-still.png differ
diff --git a/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/note-2.gif b/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/note-2.gif
new file mode 100644
index 000000000..792fda112
Binary files /dev/null and b/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/note-2.gif differ
diff --git a/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/note-2.mp4 b/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/note-2.mp4
new file mode 100644
index 000000000..27971b2e2
Binary files /dev/null and b/blog/src/assets/media/munder-difflin-vs-chatgpt-dots/note-2.mp4 differ
diff --git a/blog/src/assets/media/what-is-chatgpt-dots/lead-still.png b/blog/src/assets/media/what-is-chatgpt-dots/lead-still.png
new file mode 100644
index 000000000..1b4c7974c
Binary files /dev/null and b/blog/src/assets/media/what-is-chatgpt-dots/lead-still.png differ
diff --git a/blog/src/assets/media/what-is-chatgpt-dots/lead.gif b/blog/src/assets/media/what-is-chatgpt-dots/lead.gif
new file mode 100644
index 000000000..d83ff4064
Binary files /dev/null and b/blog/src/assets/media/what-is-chatgpt-dots/lead.gif differ
diff --git a/blog/src/assets/media/what-is-chatgpt-dots/lead.mp4 b/blog/src/assets/media/what-is-chatgpt-dots/lead.mp4
new file mode 100644
index 000000000..54f98c4bc
Binary files /dev/null and b/blog/src/assets/media/what-is-chatgpt-dots/lead.mp4 differ
diff --git a/blog/src/assets/media/what-is-chatgpt-dots/note-1-still.png b/blog/src/assets/media/what-is-chatgpt-dots/note-1-still.png
new file mode 100644
index 000000000..7323e2e28
Binary files /dev/null and b/blog/src/assets/media/what-is-chatgpt-dots/note-1-still.png differ
diff --git a/blog/src/assets/media/what-is-chatgpt-dots/note-1.gif b/blog/src/assets/media/what-is-chatgpt-dots/note-1.gif
new file mode 100644
index 000000000..b8eacb794
Binary files /dev/null and b/blog/src/assets/media/what-is-chatgpt-dots/note-1.gif differ
diff --git a/blog/src/assets/media/what-is-chatgpt-dots/note-1.mp4 b/blog/src/assets/media/what-is-chatgpt-dots/note-1.mp4
new file mode 100644
index 000000000..47183d9aa
Binary files /dev/null and b/blog/src/assets/media/what-is-chatgpt-dots/note-1.mp4 differ
diff --git a/blog/src/assets/media/what-is-chatgpt-dots/note-2-still.png b/blog/src/assets/media/what-is-chatgpt-dots/note-2-still.png
new file mode 100644
index 000000000..70b422da8
Binary files /dev/null and b/blog/src/assets/media/what-is-chatgpt-dots/note-2-still.png differ
diff --git a/blog/src/assets/media/what-is-chatgpt-dots/note-2.gif b/blog/src/assets/media/what-is-chatgpt-dots/note-2.gif
new file mode 100644
index 000000000..ecd5d1afc
Binary files /dev/null and b/blog/src/assets/media/what-is-chatgpt-dots/note-2.gif differ
diff --git a/blog/src/assets/media/what-is-chatgpt-dots/note-2.mp4 b/blog/src/assets/media/what-is-chatgpt-dots/note-2.mp4
new file mode 100644
index 000000000..8d2edc715
Binary files /dev/null and b/blog/src/assets/media/what-is-chatgpt-dots/note-2.mp4 differ
diff --git a/blog/src/posts/chatgpt-dots-alternatives.md b/blog/src/posts/chatgpt-dots-alternatives.md
new file mode 100644
index 000000000..9002acf1a
--- /dev/null
+++ b/blog/src/posts/chatgpt-dots-alternatives.md
@@ -0,0 +1,94 @@
+---
+title: "ChatGPT Dots Alternatives: 7 Always On Agents, Local and Cloud"
+seoTitle: "7 ChatGPT Dots Alternatives in 2026: Free, Local and Cloud"
+description: "Personal ChatGPT dots need Pro from $100 a month and skip the UK and EU. Seven alternatives, sorted by where they run: your computer or a vendor's cloud."
+date: 2026-09-30
+category: comparisons
+categoryLabel: Comparisons
+type: Non-technical
+primaryKeyword: "chatgpt dots alternatives"
+secondaryKeywords: ["chatgpt dots alternative", "open source chatgpt dots alternative", "free chatgpt dots alternative", "chatgpt dots not available in uk", "apps like chatgpt dots"]
+tags: ["Comparisons", "AI Agents", "Open Source", "Local-First"]
+ogImage: "https://munderdiffl.in/blog/assets/media/chatgpt-dots-alternatives/lead-still.png"
+faq:
+ - q: "Is there a free alternative to ChatGPT dots?"
+ a: "Yes. OpenClaw, Munder Difflin and Hermes Agent are free and open source and run on hardware you control. You still pay for the AI model, or use one you already subscribe to. Meta Muse has a free hosted tier, but only in the US and Canada as of 29 Sep 2026."
+ - q: "Can I use ChatGPT dots in the UK or EU?"
+ a: "Not on a personal Pro plan. At launch OpenAI excludes Pro users in the EEA, Switzerland and the UK, while Business Premium covers all supported ChatGPT regions. Gemini Spark has the same gap, so the open source picks are the simplest route there."
+ - q: "What is the best open source alternative to ChatGPT dots?"
+ a: "OpenClaw for a personal assistant in your chat apps, Munder Difflin for a team of work agents on your computer, and Hermes Agent for memory that learns over time. All three are open source and run on your own machine or server."
+ - q: "Which ChatGPT dots alternatives keep working when my laptop is off?"
+ a: "The cloud ones. Claude Cowork runs scheduled tasks on Anthropic's servers, Grok Bot works on a persistent cloud computer, and Gemini Spark might finish a task in a remote browser. The local picks need a machine that stays on."
+ - q: "Can I run more than one agent instead of a single dot?"
+ a: "Not with dots yet: at launch each person gets one primary dot. Munder Difflin runs several named agents with separate roles on your computer today, each on the CLI engine you pick."
+---
+
+The best ChatGPT dots alternatives are OpenClaw, Munder Difflin and Hermes Agent if you want agents on your own computer, and Claude Cowork, Gemini Spark, Grok Bot or Meta Muse if you want another hosted agent.
+
+
TL;DR
A free personal assistant you own: OpenClaw. A team of work agents on your computer: Munder Difflin. Memory that learns and a cron you control: Hermes Agent. Work that keeps going with your laptop shut: Claude Cowork or Grok Bot. Life in Google apps: Gemini Spark. A free hosted agent in the US or Canada: Meta Muse.
+
+Three run on your computer, four in a cloud.
+
+New to dots? Read [What is ChatGPT dots?](/blog/what-is-chatgpt-dots/).
+
+## Why look for a ChatGPT dots alternative?
+
+Dots are pricey, regional and single at launch. From [OpenAI's launch post](https://openai.com/index/introducing-dots/) and help pages, as of 30 Sep 2026:
+
+- **Price.** Your first dot comes with ChatGPT Pro, from $100 a month, or Business Premium. Free, Go and Plus do not get one.
+- **Region.** Pro users in the EEA, Switzerland and the UK are left out at launch.
+- **One dot.** You get one primary dot. OpenAI writes that "over time, we envision teams of dots working together", which is a plan, not a feature.
+- **Where the work lives.** A dot works on its own cloud computer at OpenAI. Access to your own computer is optional and off by default.
+
+Dots still have real strengths: zero setup, a mobile app, voice calls and over 4,000 app plugins.
+
+## Does it run on your computer or in the cloud?
+
+This settles who holds your files and whether work stops when you close the lid. Three picks run on hardware you control, four on a vendor's cloud, like dots. We drew the same line in our [Grok Bot alternatives](/blog/grok-bot-alternatives/) guide, and [why local first matters for AI agents](/blog/why-local-first-matters-for-ai-agents/) explains which side we lean to.
+
+
+
+### Runs on your computer
+
+**1. OpenClaw.** [OpenClaw](https://github.com/openclaw/openclaw) is an open source (MIT) personal assistant that runs on your own computer and answers in WhatsApp, Telegram, Discord, Slack, Teams, iMessage and 20+ more channels. It runs as a background daemon, keeps memory and credentials on your hardware, and has no paid tier. Best for: a personal errand runner you own.
+
+**2. Munder Difflin.** [Munder Difflin](https://harnessmd.com/download) is what we make: a free and open source desktop app that runs a team of named AI agents on your own computer. Each agent has a role and runs a CLI engine you already use: Claude Code, Codex, Gemini CLI, OpenCode, Cursor and more. Each one keeps its own `memory.md`, inbox and outbox. You can put missions on a schedule, and a Slack channel can feed work straight to Michael, your lead agent. The catch: agents run while your computer runs, and you install the CLIs yourself. Best for: engineering and project work across several engines. (Munder Difflin 0.5.3, checked 30 Sep 2026.)
+
+**3. Hermes Agent.** [Hermes Agent](https://github.com/NousResearch/hermes-agent) from Nous Research is open source (MIT) and built around a learning loop: it curates its own memory and writes new skills from experience. Its cron scheduler runs jobs unattended, and you reach it over Telegram, Discord, Slack, WhatsApp or Signal. Best for: memory that grows, on a box you control.
+
+### Runs in the cloud
+
+**4. Claude Cowork.** Cowork is Anthropic's agent for longer pieces of work, and it now runs in the cloud. Anthropic's [Cowork help page](https://support.claude.com/en/articles/13345190-get-started-with-cowork) says "Claude's work runs on Anthropic's servers, in an isolated environment" and that scheduled tasks don't need your computer to be awake. Since 16 Sep 2026 Anthropic has been [folding Cowork into the main Claude app](https://claude.com/blog/cowork-is-now-claude), Pro and Max first. Claude Pro was $20 a month, billed monthly, on 30 Sep 2026. Best for: reports, files and research, task by task.
+
+**5. Gemini Spark.** Spark is Google's personal agent in the Gemini app. It works from your Connected Apps and chats, and can drive your local Chrome. If your device goes off before a task is done, Google's [Spark help page](https://support.google.com/gemini/answer/17094507) says it "might use a remote browser" in the cloud instead. You need Google AI Pro or Ultra and to be 18 or over. Like dots on Pro, it is not available in the EEA, Switzerland or the UK, nor in Nigeria. Best for: Google app users outside Europe.
+
+**6. Grok Bot.** SpaceXAI's docs say your Bots run on "a persistent cloud computer with a browser, filesystem, and terminal", and closing your laptop or phone does not stop a routine. It comes with paid Cursor plans or a SuperGrok account. Cursor's Individual plan was $20 a month on 30 Sep 2026. [Grok Bot pricing](/blog/grok-bot-pricing/) has the other plans. Best for: hosted work agents whose computer never sleeps.
+
+**7. Meta Muse.** [Meta's help centre](https://www.meta.com/help/artificial-intelligence/1047255454427887/) says Muse runs on "a cloud-based environment called a virtual machine". On 29 Sep 2026 Meta said Muse is "available in the US and Canada" and free for most use; its [subscriptions page](https://www.meta.com/help/subscriptions/1021145227643680/) lists Power at $20 a month and Max at $100. Best for: a free hosted agent in North America. Elsewhere, see [Meta Muse alternatives](/blog/meta-muse-alternatives/).
+
+## ChatGPT dots alternatives, checked 30 Sep 2026
+
+| Agent | Runs on | Always on or task based | Who can get it | Price to start |
+| --- | --- | --- | --- | --- |
+| ChatGPT dots | OpenAI's cloud computer | Always on | Pro outside EEA, Switzerland, UK; Business Premium | Pro, $100/month. |
+| OpenClaw | Your computer | Always on daemon | Anyone | Free and open source. |
+| Munder Difflin | Your computer | Runs while your computer does, plus schedules | Anyone on Mac, Windows, Linux | Free and open source. |
+| Hermes Agent | Your computer or your server | Always on, cron jobs | Anyone | Free and open source. |
+| Claude Cowork | Anthropic's cloud | Tasks, plus scheduled tasks | Paid Claude plans | Claude Pro, $20/month. |
+| Gemini Spark | Google's cloud and your Chrome | Ongoing tasks and schedules | Google AI Pro or Ultra; not EEA, Switzerland, UK, Nigeria | Google AI Pro, price by country. |
+| Grok Bot | SpaceXAI's cloud computer | Always on, routines | Paid Cursor plans or SuperGrok | Cursor Individual, $20/month. |
+| Meta Muse | Meta's cloud VM | Tasks | US and Canada | Free tier. |
+
+## Is there a free alternative to ChatGPT dots?
+
+Yes: OpenClaw, Munder Difflin and Hermes Agent are all free and open source. The software costs nothing, but the model still does. OpenClaw and Hermes Agent take API keys, local models or a provider login. Munder Difflin drives the CLIs you already have, so the Claude or ChatGPT subscription you pay for today does the work.
+
+
+
+Meta Muse is the only hosted pick with a free tier, and only in the US and Canada. Dots start at $100 a month on Pro (30 Sep 2026), and OpenAI has not said how dots usage will count against your plan after the first month.
+
+## What is the best open source alternative to ChatGPT dots?
+
+OpenClaw, if you want what a dot does for your personal life: one assistant, in your chat apps, on your hardware. Munder Difflin, if you want agents for work: several named agents with roles, each on the engine that suits the job. Hermes Agent is strongest on memory.
+
+None of the three is a hosted computer that works with your laptop shut: you need a server, or a machine that stays on. We put Munder Difflin and dots side by side in [Munder Difflin vs ChatGPT dots](/blog/munder-difflin-vs-chatgpt-dots/).
diff --git a/blog/src/posts/github-copilot-cli-vs-claude-code-for-hive-work.md b/blog/src/posts/github-copilot-cli-vs-claude-code-for-hive-work.md
index 454436a09..1113ab1c1 100644
--- a/blog/src/posts/github-copilot-cli-vs-claude-code-for-hive-work.md
+++ b/blog/src/posts/github-copilot-cli-vs-claude-code-for-hive-work.md
@@ -1,103 +1,120 @@
---
-title: "GitHub Copilot CLI vs Claude Code as Hive Workers"
-description: "How GitHub Copilot CLI and Claude Code behave as agent engines inside a multi-agent harness: hooks, inbox mail, orchestration, session lifecycle — and why the right answer is to run both."
+title: "Claude Code vs GitHub Copilot: CLI, Editor and Pricing Compared"
+description: "Claude Code vs GitHub Copilot, checked 29 Sep 2026: Copilot CLI and the editor, AI Credits vs Claude plans, models, hooks, and when to pick each."
date: 2026-07-03
+updated: 2026-09-29
category: comparisons
categoryLabel: Comparisons
type: Technical
-primaryKeyword: "github copilot cli vs claude code"
-secondaryKeywords: ["copilot cli multi-agent", "claude code multi-agent harness", "copilot cli print mode", "agent engine comparison", "copilot cli automation", "claude code hooks"]
-tags: ["Comparisons", "Claude Code", "Copilot CLI", "Multi-Agent", "Engines"]
+primaryKeyword: "claude code vs github copilot"
+secondaryKeywords: ["github copilot vs claude code", "copilot cli vs claude code", "github copilot cli vs claude code", "claude code vs github copilot pricing", "is claude code better than copilot"]
+tags: ["Comparisons", "Claude Code", "Copilot CLI", "GitHub Copilot", "Engines"]
author:
name: Chaitanya Giri
initials: CG
faq:
- - q: "Can GitHub Copilot CLI work as an agent in a multi-agent system?"
- a: "Yes, with a specific shape. Copilot CLI's documented programmatic mode (copilot -p) executes one prompt and exits after completion, which makes it excellent for dispatched, self-contained tasks: hand it a scoped job, it runs autonomously with --allow-all-tools and --no-ask-user, returns a result, and terminates. What it can't do in that mode is stay resident and react to incoming messages, because there is no long-lived process to deliver them to."
- - q: "Why is Claude Code better suited to long-lived hive work?"
- a: "Claude Code exposes lifecycle hooks that an external harness can attach to. In Munder Difflin, those hooks let the harness see every turn, gate actions for human approval, steer mid-run, and — critically — drain the agent's inbox at the end of each turn via the Stop hook, so a Claude worker picks up new mail without being respawned. That persistent, observable lifecycle is what makes it eligible to be a worker that converses, or the GOD orchestrator itself."
- - q: "What happens when you send a message to a Copilot CLI worker in Munder Difflin?"
- a: "It doesn't silently disappear — it bounces to the GOD orchestrator. Because Copilot's print mode exits per turn, there's no resident process with a hook bridge to drain an inbox, so the provider is registered with canReceiveInbox set to false and routed mail is redirected to the orchestrator, which can re-dispatch the work as a fresh Copilot task or hand it to a hive-aware worker."
- - q: "Does GitHub Copilot CLI have hooks of its own?"
- a: "Yes. Copilot CLI supports hooks defined in JSON files (in .github/hooks/ per repo or ~/.copilot/hooks/ personally), including sessionStart and sessionEnd events, plus MCP servers and custom instructions. Munder Difflin's v0.3.3 integration doesn't bridge those into the hive yet — it drives the documented print mode, where the process exits after each prompt — so within the harness a Copilot worker is non-hive-aware by design."
- - q: "How is Copilot CLI billed compared to Claude Code?"
- a: "Copilot CLI authenticates with your existing GitHub Copilot login and consumes premium requests from your Copilot plan — GitHub documents one premium request per prompt on the default model, multiplied by the model's rate for others. Claude Code runs on your Anthropic plan or API key. Practically, that means a mixed hive spreads work across two subscriptions you may already be paying for."
- - q: "Should I pick one engine or mix them?"
- a: "Mix them. Use Claude Code for the orchestrator and long-lived workers that need to receive mail, converse across turns, and run under hook-driven guardrails. Use Copilot CLI workers for burst work: scoped, self-contained tasks dispatched by the orchestrator that finish in one shot. Munder Difflin lets you choose the engine per hire, so this is a per-agent decision, not a platform decision."
+ - q: "Is Claude Code better than GitHub Copilot?"
+ a: "For long, multi-file agent work that you hand off and review, Claude Code is usually the stronger tool, because its hooks, permission modes and background sessions go further. For as-you-type completions, IDE coverage and models from several labs on one bill, Copilot is better value. Many developers keep Copilot in the editor and use Claude Code for bigger tasks."
+ - q: "Is Copilot CLI the same as Copilot in VS Code?"
+ a: "No. They are separate clients on the same Copilot subscription. Copilot CLI is a terminal agent you start with the copilot command, much like claude, while Copilot in VS Code adds completions, chat and agent mode inside the editor. Both draw on the same monthly AI Credits for chat and agent work."
+ - q: "Can GitHub Copilot use Claude models?"
+ a: "Yes. On 29 Sep 2026 GitHub's supported models page listed Claude Sonnet 5.5, Opus 5.5 and Fable 5.1, plus older Claude models, in Copilot Chat and Copilot CLI. You choose them with /model or --model, and each one spends AI Credits at its own token rate. That gives you Claude models, not the Claude Code tool itself."
+ - q: "Does Copilot still bill premium requests?"
+ a: "Not on the current billing platform. GitHub moved Copilot to usage-based billing on 1 June 2026, so chat, agent and CLI use now spends AI Credits worth one US cent each, priced by tokens. Copilot CLI 1.0.89's own billing help still mentions premium requests, but only for accounts on the legacy platform."
+ - q: "Does Copilot CLI read CLAUDE.md?"
+ a: "Yes. GitHub's custom instructions docs list `AGENTS.md`, `CLAUDE.md` (also `.claude/CLAUDE.md`) and `GEMINI.md` among the files Copilot CLI loads, next to `.github/copilot-instructions.md`. A repo already set up for Claude Code gives Copilot CLI the same project rules without a second file."
---
-
TL;DR
Inside a multi-agent harness, the engine question isn't "which CLI writes better code" — it's what lifecycle the process exposes. Claude Code runs as a long-lived session with native hooks, so Munder Difflin can gate it, steer it, drain its inbox on the Stop hook, and even make it the GOD orchestrator. GitHub Copilot CLI (new in v0.3.3) runs in its documented print mode — copilot -p — which exits after every prompt: brilliant for dispatched, self-contained tasks on your existing Copilot subscription, but with no resident process to receive mail, routed messages bounce to the orchestrator instead. Verdict: mix them — Claude as orchestrator and long-lived workers, Copilot for burst tasks.
+Claude Code is the stronger choice for long, multi-file agent work driven from the terminal. GitHub Copilot is the better value if you want completions as you type, models from several labs and GitHub-native agents on one bill. Plenty of developers pay for both.
-Munder Difflin v0.3.3 added GitHub Copilot CLI as its seventh agent engine — the project's first community-contributed provider ([PR #101](https://github.com/chaitanyagiri/munder-difflin/pull/101) by @anxkhn). That makes a comparison suddenly practical rather than theoretical: you can hire a Claude Code worker and a Copilot worker onto the same floor and watch them behave differently. This post is about *that* difference — not "which model is smarter," but how each CLI behaves as a **worker inside a hive**.
+By Copilot, people mean either the Copilot CLI (`copilot`, a terminal agent much like `claude`) or Copilot inside VS Code and JetBrains. This page covers both, checked on 29 Sep 2026. Both teams ship so often that a comparison table has the shelf life of an open carton of milk, hence the date.
-Both are genuinely good tools. The differences below are lifecycle differences, and lifecycle is what a harness cares about.
+## What is the difference between Claude Code and GitHub Copilot?
-## The thing a harness actually needs from an engine
+Claude Code is an agent first, and GitHub Copilot is a platform. Claude Code takes a task, reads the repo, edits files, runs commands and checks its own work, using Anthropic's models only. If you are on a Max plan, [our Max plan guide](/blog/claude-code-max-plan-tips/) covers how to make that usage last.
-A [multi-agent harness](/blog/what-is-a-multi-agent-harness/) wraps real CLI processes and coordinates them: routing messages between inboxes, gating destructive actions, tracking turns, waking idle agents when mail arrives. To do all that, it needs answers from the engine to three questions:
+Copilot started as autocomplete and grew outwards: completions and next edit suggestions, chat and agent mode, the Copilot CLI, a cloud agent that opens pull requests, and a model picker spanning OpenAI, Anthropic, Google, xAI and others.
-1. **Does the process stay alive between turns?** (Can it hold a conversation, or is every invocation a fresh start?)
-2. **Does it emit lifecycle signals a harness can attach to?** (Turn ended, tool about to run, session stopped?)
-3. **Can it be interrupted, steered, and fed new input mid-stream?**
+If you write most of the code yourself and want help line by line, Copilot is built for that. If you hand off whole tasks and review diffs, Claude Code is.
-Claude Code and Copilot CLI give very different answers — and neither answer is wrong. They're different shapes for different jobs.
+## Claude Code vs GitHub Copilot at a glance
+
+We ran both version commands on macOS on 29 Sep 2026:
+
+```
+$ claude --version
+2.1.284 (Claude Code)
+$ copilot --version
+GitHub Copilot CLI 1.0.89.
+Run 'copilot update' to check for updates.
+```
{% img "note-1" %}
-## Claude Code: the hive-aware resident
+| | Claude Code | GitHub Copilot |
+|---|---|---|
+| Version checked | 2.1.284 | CLI 1.0.89 |
+| Where it runs | Terminal, VS Code, JetBrains, desktop app, web | VS Code, Visual Studio, JetBrains, Xcode, Eclipse, GitHub.com, terminal |
+| Completions as you type | No | Yes, unlimited on paid plans |
+| Models | Claude only (Fable, Opus, Sonnet, Haiku by plan) | OpenAI, Anthropic, Google, xAI, Moonshot, Microsoft, or `auto` |
+| Project instructions | `CLAUDE.md`, can also read `AGENTS.md` | `.github/copilot-instructions.md`, `AGENTS.md`, `CLAUDE.md`, `GEMINI.md` |
+| Hooks | 33 events, such as PreToolUse and Stop | 14 events, such as preToolUse and agentStop |
+| Headless run | `claude -p` | `copilot -p` |
+| Parallel work | Subagents, background sessions (`--bg`) | Subagents, `--fleet` |
-Claude Code runs as a persistent interactive session in its pseudo-terminal, and it exposes native hooks — external commands fired at lifecycle points. Munder Difflin leans on those hard:
+## Copilot CLI vs Claude Code: how do the terminal agents compare?
-- **Turn-end inbox drain.** When a Claude worker finishes a turn, the Stop hook fires and the harness checks the agent's mailbox. If mail is waiting, the agent drains it *in the same session* — no respawn, no lost context. This is what makes a worker conversational: other agents can message it and it will actually respond.
-- **Hook-driven guardrails.** The human-in-the-loop gate, mid-run steering, and graceful stop are all driven through Claude Code hook returns — you can intervene without killing the session.
-- **Orchestrator eligibility.** Because the process is long-lived and observable, Claude Code can be the [GOD orchestrator](/blog/how-the-god-orchestrator-works/) itself — the always-on supervisor that adjudicates traffic and routes tasks.
+Closely: both have a plan mode, a headless `-p` mode, MCP servers, skills, custom agents, hooks and session resume. The differences, read from each tool's `--help` on 29 Sep 2026:
-If you want the deeper treatment of driving Claude Code as a coordinated fleet, see the [Claude Code orchestration guide](/blog/claude-code-orchestration-guide/).
+* **GitHub is built in on the Copilot side.** Copilot CLI ships GitHub's MCP server by default (`--disable-builtin-mcps` turns it off), so issues, pull requests and Actions need no setup. Claude Code reaches GitHub through `gh` or an MCP server you add.
+* **Spending caps differ.** Copilot CLI has `--max-ai-credits` per session, and its autopilot mode (`--mode autopilot`) pauses after 5 automatic continuations unless you raise `--max-autopilot-continues`. Claude Code's closest cap is `--max-budget-usd`, which only works with `--print`.
+* **Parallelism differs.** `copilot --fleet` runs one prompt as parallel subagents. Claude Code has subagents too, and `claude --bg` starts a background session that `claude attach` picks up later.
+* **Model choice differs.** `copilot --model auto` lets Copilot pick the model. `claude --model` takes aliases such as `opus`, `sonnet` and `fable`; on Pro, Max, Team and Enterprise the default is Opus 5.5, per [Anthropic's model docs](https://code.claude.com/docs/en/model-config).
+* **Permissions are shaped differently.** Copilot uses `--allow-tool`, `--deny-tool` and `--allow-all` (also spelled `--yolo`). Claude Code uses permission modes plus `--allowedTools` and `--disallowedTools`.
-## GitHub Copilot CLI: the burst specialist
+Copilot CLI can run the same Claude models, so the harness is the dividing line: Claude Code has the wider hook surface and first class background sessions, Copilot CLI has GitHub itself. Per [GitHub's custom instructions docs](https://docs.github.com/en/copilot/how-tos/copilot-cli/customize-copilot/add-custom-instructions), Copilot CLI also reads `CLAUDE.md`, so switching does not mean rewriting your project rules.
-GitHub Copilot CLI is GitHub's terminal agent, and its docs describe two modes: an interactive session, and a **programmatic mode** built for scripting — per GitHub's own reference, `-p` will "execute a prompt programmatically (exits after completion)." That exit-per-turn behavior is the documented design, and Munder Difflin drives it exactly as documented. The spawn shape from the v0.3.3 changelog:
+## Is Claude Code better than Copilot in VS Code?
-```
-copilot -p "" -s --allow-all-tools --no-ask-user [--model ]
-```
+Not for typing help: Copilot gives completions and next edit suggestions as you type, which spend no AI Credits on paid plans, and Claude Code has none. Its VS Code and JetBrains extensions are a panel for the agent, with inline diffs, @ mentions, plan review and history, running the same engine as the CLI.
-Piece by piece: `-p` passes the task, `-s` (silent) outputs only the agent response, `--allow-all-tools` auto-approves tool use (GitHub notes it's required for programmatic runs — and in Munder Difflin these auto-approval flags are gated by the floor's auto-mode toggle, like every other engine), and `--no-ask-user` disables clarifying questions so the run is fully autonomous. There's a model picker (Claude Sonnet 4.5 by default, GPT-5.4, or `auto`) and best-effort `--resume` session continuity.
+Copilot's agent mode covers similar ground to Claude Code's panel, with any model on your plan. The two coexist in one editor: Copilot for completions, Claude Code for the tasks you hand off.
-Two things make Copilot workers genuinely attractive:
+## How much do Claude Code and GitHub Copilot cost?
-- **It's on your existing subscription.** Copilot CLI authenticates with your GitHub Copilot login — no new API key. GitHub bills it in premium requests (one per prompt on the default model, times the model's multiplier), drawing from the allowance you may already have.
-- **Exit-per-turn is a feature for dispatched work.** A scoped task arrives, the process runs it to completion with no residual state, and terminates cleanly. For burst tasks, that's exactly what you want.
+As of 29 Sep 2026, Copilot is cheaper to start: free, then Pro at $10 a month, while Claude Code needs Claude Pro at $20 a month ($17 yearly) or API credit.
-The honest limitation: with no resident process between turns, there's nothing to deliver inbox mail *to*. So in Munder Difflin, Copilot's provider is registered non-hive-aware (`canReceiveInbox: false`), and routed mail **bounces to the GOD orchestrator** instead of silently dropping. The orchestrator can then re-dispatch it as a fresh Copilot task or hand it to a hive-aware worker.
+| Checked 29 Sep 2026 | Claude Code | GitHub Copilot |
+|---|---|---|
+| Free option | None. The Free Claude plan does not include Claude Code | Copilot Free: 2,000 completions a month, limited chat and agent use, Copilot CLI included |
+| Cheapest paid individual plan | Pro: $20 a month, or $17 a month billed yearly ($200 up front) | Pro: $10 a month. No yearly price listed |
+| Top individual plan | Max 20x: $200 a month (Max 5x is $100). Monthly only | Max: $100 a month (Pro+ is $39). No yearly price listed |
+| Team or business seat | Team Standard: $25 a seat a month, or $20 billed yearly. Team Premium: $125 a seat, or $100 billed yearly | Business: $19 a seat a month. No yearly price listed |
+| Enterprise | $20 a seat a month, billed yearly only, plus usage at API rates | $39 a seat a month, on GitHub Enterprise Cloud. No yearly price listed |
+| What the plan includes | Usage limits, not credits: a five-hour window plus weekly limits, shared with Claude chat. Max gives 5x or 20x Pro's usage per session, a Premium seat 5x a Standard seat. Anthropic publishes no credit or dollar figure for these limits. Usage based Enterprise includes no usage: all of it bills at API rates from the first token | AI Credits (since 1 June 2026) at one cent each, priced by tokens: Pro 1,500 (1,000 base plus 500 flex), Pro+ 7,000 (3,900 plus 3,100), Max 20,000 (10,000 plus 10,000). Business 1,900 and Enterprise 3,900 a seat, pooled across the organisation. Flex can change. Completions unlimited on paid plans |
+| What happens at the limit | Wait for the five-hour or weekly reset, or move up a plan. On Pro and Max, turn on usage credits: prepaid, at API rates, under a monthly spending cap you set or no cap. On Team and seat based Enterprise, usage credits stay off until an Owner turns them on (Team prepays, Enterprise is billed monthly), and Owners set org and user spend limits | Wait for the reset at 00:00 UTC on the 1st, or upgrade. Individuals keep working only after they set a dollar budget, and extra credits bill at one cent each. On Business and Enterprise, paid usage past the pool is on by default; admins set budgets or turn it off, and Copilot then pauses until the 1st |
+| Pay per use | API credit through a Claude Console account, billed per token at API rates, no plan needed | Extra AI Credits at one cent each on top of a paid plan. Free cannot buy extra credits |
-To be fair to Copilot CLI: the tool itself is more extensible than the print-mode integration uses. GitHub ships its own hooks system (JSON files in `.github/hooks/` or `~/.copilot/hooks/`, with `sessionStart`/`sessionEnd` events), MCP server support, and custom instructions. A deeper bridge is conceivable; v0.3.3 ships the documented, honest version.
+Sources, read 29 Sep 2026: Claude's [pricing page](https://claude.com/pricing), [Max plan](https://support.claude.com/en/articles/11049741-what-is-the-max-plan), [usage credits](https://support.claude.com/en/articles/12429409-extra-usage-for-paid-claude-plans), [Team and Enterprise extra usage](https://support.claude.com/en/articles/12005970) and [setup](https://code.claude.com/docs/en/setup) docs; GitHub's [plans page](https://github.com/features/copilot/plans), [plans docs](https://docs.github.com/en/copilot/get-started/plans), the [1 June billing changelog](https://github.blog/changelog/2026-06-01-updates-to-github-copilot-billing-and-plans) and billing docs for [individuals](https://docs.github.com/en/copilot/concepts/billing-and-usage/individuals/billing) and [organisations](https://docs.github.com/en/copilot/concepts/billing-and-usage/organizations-and-enterprises/billing).
-{% img "note-2" %}
+The practical difference is the meter. Copilot gives you a monthly pot of credits, then bills a cent a credit against your budget or waits for the 1st. Claude Code gives you five-hour and weekly limits that reset on their own, then bills API rates only if you turn on usage credits. Copilot's cheapest plan, top individual plan and business seat all cost less (Claude's Enterprise seat is cheaper, $20 against $39, but every token is billed on top), and only Claude lists a yearly price.
-## Verdict: don't pick — mix
+## Can Copilot use Claude models or Claude Code?
-The comparison table writes itself once you frame it as lifecycle:
+Copilot can use Claude models and hand work to a Claude agent on GitHub, but neither is the Claude Code CLI on your machine. GitHub's supported models page, checked 29 Sep 2026, lists Claude Sonnet 5.5, Opus 5.5 and Fable 5.1 in Copilot Chat and Copilot CLI. You can also assign an issue to Anthropic's Claude agent on GitHub.com, which works in the cloud and opens a pull request; GitHub marks that as public preview and bills it in AI Credits plus Actions minutes. Running `claude` locally still takes a Claude plan or API key.
-| | Claude Code | GitHub Copilot CLI |
-|---|---|---|
-| Process lifetime | Long-lived session | Exits after each prompt (`-p`) |
-| Harness hook bridge | Native hooks (incl. Stop-hook inbox drain) | None in print mode |
-| Receives hive mail | Yes, drains inbox in-session | No — bounces to orchestrator |
-| Orchestrator-eligible | Yes | No |
-| Auth | Anthropic plan / API key | Existing GitHub Copilot login |
-| Sweet spot | Orchestrator, conversational workers | Dispatched, self-contained burst tasks |
+## Can you run Claude Code and Copilot CLI side by side?
-Because Munder Difflin makes the engine a per-hire choice, this isn't an either/or. The pattern that works: **Claude Code as the orchestrator and your long-lived workers** — the agents that need to receive messages, hold context across turns, and run under hook-driven approval gates — and **Copilot CLI workers for burst tasks** the orchestrator dispatches: a scoped refactor, a test-writing pass, a one-shot investigation. Different engines for different roles is the general lesson of mixed fleets, and it also spreads cost across two subscriptions you likely already pay for.
+Yes, and because Copilot CLI reads `CLAUDE.md`, both agents see the same project rules. You can do it by hand with two terminals and a git worktree each, or use [Munder Difflin](https://harnessmd.com/download), free and open source, which hires both as agents on one floor.
+
+Checked against tag v0.5.3 of the app repo: `src/shared/agentProvider.ts` lists twelve CLIs, Claude Code and Copilot among them. Claude Code can run Michael, the orchestrator, and takes messages between turns. Copilot is driven in print mode, `copilot -p "" -s --allow-all-tools --no-ask-user`, with the autonomy flags added only when the floor's auto mode is on. That process exits after each prompt, so the Copilot entry sets `canReceiveInbox: false`: mail for a Copilot worker goes to Michael instead, and `modelProvidersForAgent` in `src/renderer/src/store/config.ts` keeps Copilot off the orchestrator list. That is a real limit, so give Copilot workers self-contained tasks. [The Copilot agent setup guide](/blog/how-to-add-a-github-copilot-cli-agent/) walks through hiring one, and [what a multi-agent harness is](/blog/what-is-a-multi-agent-harness/) explains the rest of the floor.
+
+{% img "note-2" %}
-For everything else in the release — including the built-in Monaco IDE — see the [v0.3.3 launch post](/blog/launching-munder-difflin-v0-3-3/).
+## Which should you pick: Claude Code or GitHub Copilot?
-## Try both on one floor
+**Pick Claude Code if** you hand off whole tasks and review diffs, your changes span many files, you want hooks and background sessions, or you already pay for Claude Pro or Max.
-Hire a Claude worker and a Copilot worker side by side and watch the difference yourself — [download Munder Difflin](https://github.com/chaitanyagiri/munder-difflin/releases/latest) (free, MIT, local-first), and if it's useful, [a GitHub star](https://github.com/chaitanyagiri/munder-difflin) helps more people find it.
+**Pick GitHub Copilot if** you want completions as you type, several labs' models on one bill, your work lives in GitHub issues and pull requests, or you want a free tier and a $10 plan.
-Sources: [GitHub Docs — Copilot CLI command reference](https://docs.github.com/en/copilot/reference/copilot-cli-reference/cli-command-reference);
-[GitHub Docs — Running Copilot CLI programmatically](https://docs.github.com/en/copilot/how-tos/copilot-cli/automate-copilot-cli/run-cli-programmatically);
-[GitHub Docs — About Copilot CLI](https://docs.github.com/en/copilot/concepts/agents/about-copilot-cli);
-[GitHub Docs — Hooks for Copilot](https://docs.github.com/en/copilot/concepts/agents/hooks);
-[GitHub Docs — Copilot premium requests](https://docs.github.com/en/copilot/concepts/billing/copilot-requests).
+**Pick both if** you want Copilot's completions and Claude Code for agent work. Copilot Pro plus Claude Pro came to $30 a month on 29 Sep 2026.
diff --git a/blog/src/posts/grok-bot-pricing.md b/blog/src/posts/grok-bot-pricing.md
index 4f4db8bcb..768c6ba3e 100644
--- a/blog/src/posts/grok-bot-pricing.md
+++ b/blog/src/posts/grok-bot-pricing.md
@@ -32,7 +32,7 @@ Grok Bot is SpaceXAI's team of always on AI agents, called Bots. They sign into
If what you want is agents for your work, there is also a way to pay nothing for the app. [Munder Difflin](https://harnessmd.com/download) runs a team of agents on your own computer using AI engines you already have. It is free and open source. We compare the costs at the end.
-Every plan with Grok Bot, then Munder Difflin: a free app that runs on the AI engine you already use.
+Every plan with Grok Bot, then Munder Difflin: a free app that runs on the AI engine you already use.
diff --git a/blog/src/posts/long-running-agents-the-2026-shift.md b/blog/src/posts/long-running-agents-the-2026-shift.md
index 6039fe3f9..e69da8b5a 100644
--- a/blog/src/posts/long-running-agents-the-2026-shift.md
+++ b/blog/src/posts/long-running-agents-the-2026-shift.md
@@ -37,6 +37,8 @@ biggest shift of 2026 is that this is no longer the ceiling. Agents now **run fo
goal and working toward it across hundreds of steps before they need you again. That changes what an
agent can *do*, and it changes what you have to *build* to make it work.
+OpenAI's [ChatGPT dots](/blog/what-is-chatgpt-dots/), launched on 29 Sep 2026, push the idea one step further: an agent that keeps a standing responsibility instead of one goal.
+
## The trend, in numbers
This isn't a vibe; it's measurable. [METR](https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/)
diff --git a/blog/src/posts/meta-muse-alternatives.md b/blog/src/posts/meta-muse-alternatives.md
index 52bef65b3..fb072608c 100644
--- a/blog/src/posts/meta-muse-alternatives.md
+++ b/blog/src/posts/meta-muse-alternatives.md
@@ -36,7 +36,7 @@ Plenty of people still can't use it, or don't want to. There are three common re
Here are seven alternatives, and which reason each one solves.
-Muse is US only for now. Three of the seven alternatives run on your own computer.
+Muse is US only for now. Three of the seven alternatives run on your own computer.
@@ -70,6 +70,8 @@ Each agent gets a desk, an inbox, memory in plain markdown files and a shared ta
**Solves:** all three reasons, if your need is work rather than errands.
**The catch:** agents run while your computer runs. For overnight work, keep a machine awake, like a [Mac mini on your desk](/blog/run-munder-difflin-on-a-mac-mini/).
+For how it compares with OpenAI's new always on agents, read [ChatGPT dots vs Munder Difflin](/blog/munder-difflin-vs-chatgpt-dots/).
+
## 3. ChatGPT Work: the hosted pick for work
ChatGPT Work is OpenAI's agent inside ChatGPT. Plugins connect it to Slack, Microsoft Teams, Google Drive, email and calendars. Scheduled Tasks run on a timer or when something happens, and you decide what needs your approval. In the Mac and Windows desktop app it can also work with your local files. It is available in the countries OpenAI supports.
@@ -77,6 +79,8 @@ ChatGPT Work is OpenAI's agent inside ChatGPT. Plugins connect it to Slack, Micr
**Solves:** the US only limit.
**The catch:** your work runs through OpenAI's cloud, and you need a ChatGPT plan that includes it.
+OpenAI added always on agents called dots on 29 Sep 2026, on Pro and Business Premium only. If those are out of reach too, see [ChatGPT dots alternatives](/blog/chatgpt-dots-alternatives/).
+
## 4. Gemini Spark: for people who live in Google
Gemini Spark suits you if your life already runs on Gmail, Drive and Calendar. Google says it keeps working with your phone and laptop off and checks with you before major actions. It comes with Google AI Pro and Ultra in select countries, plus select business users.
diff --git a/blog/src/posts/meta-muse-grok-bot-privacy.md b/blog/src/posts/meta-muse-grok-bot-privacy.md
index 1ad177972..da76ca3ae 100644
--- a/blog/src/posts/meta-muse-grok-bot-privacy.md
+++ b/blog/src/posts/meta-muse-grok-bot-privacy.md
@@ -30,7 +30,7 @@ A personal AI agent is only useful if it can act for you. To act for you, it nee
We read the security pages of both, [Meta's Muse privacy page](https://www.meta.com/help/artificial-intelligence/1047255454427887/) and [SpaceXAI's Grok Bot security FAQ](https://docs.x.ai/grok-bot/security-faq), and put the answers side by side. We also cover the third option: [Munder Difflin](https://harnessmd.com/download), which runs a team of work agents on your own computer so the logins never leave it. It is free and open source.
-Where your logins live: in Meta's vault, on one computer shared by all your Bots, or on your own machine.
+Where your logins live: in Meta's vault, on one computer shared by all your Bots, or on your own machine.
diff --git a/blog/src/posts/meta-muse-vs-grok-bot.md b/blog/src/posts/meta-muse-vs-grok-bot.md
index 67ce24b28..6e8d29a9d 100644
--- a/blog/src/posts/meta-muse-vs-grok-bot.md
+++ b/blog/src/posts/meta-muse-vs-grok-bot.md
@@ -30,7 +30,7 @@ Meta launched Muse on 8 September 2026. SpaceXAI put Grok Bot into beta on 11 Au
You can also skip the rented computer. [Munder Difflin](https://harnessmd.com/download) runs a whole office of agents on your own machine with the AI engines you already pay for. It is free and open source. It is built for work, not errands, and we cover where it fits below.
-Muse gives every person a separate computer with a guard at the door. Grok Bot gives all your Bots one shared computer. Munder Difflin uses the computer you already own.
+Muse gives every person a separate computer with a guard at the door. Grok Bot gives all your Bots one shared computer. Munder Difflin uses the computer you already own.
diff --git a/blog/src/posts/munder-difflin-vs-chatgpt-dots.md b/blog/src/posts/munder-difflin-vs-chatgpt-dots.md
new file mode 100644
index 000000000..bbdf5f192
--- /dev/null
+++ b/blog/src/posts/munder-difflin-vs-chatgpt-dots.md
@@ -0,0 +1,104 @@
+---
+title: "ChatGPT Dots vs Munder Difflin: One Cloud Agent or a Team on Your Computer"
+description: "ChatGPT dots vs Munder Difflin: one hosted agent on OpenAI's cloud, or a free, open source team of agents on your own computer. Checked 30 Sep 2026."
+date: 2026-09-30
+category: comparisons
+categoryLabel: Comparisons
+type: Non-technical
+primaryKeyword: "chatgpt dots vs munder difflin"
+secondaryKeywords: ["munder difflin vs chatgpt dots", "chatgpt dots open source alternative", "chatgpt dots local", "always on ai agents on your own computer", "chatgpt dots alternative"]
+tags: ["Comparisons", "AI Agents", "Local-First", "Multi-Agent", "Open Source"]
+ogImage: "https://munderdiffl.in/blog/assets/media/munder-difflin-vs-chatgpt-dots/lead-still.png"
+faq:
+ - q: "What is the difference between ChatGPT dots and Munder Difflin?"
+ a: "A dot is one always on agent that OpenAI hosts on its own cloud computer, powered by GPT-6 Astra and reached through ChatGPT, Slack, Teams or a voice call. Munder Difflin is a free and open source desktop app that runs a team of named agents on your own computer, each one a CLI engine you choose, such as Claude Code, Codex or Gemini CLI."
+ - q: "Is there an open source alternative to ChatGPT dots?"
+ a: "Yes. Munder Difflin is MIT licensed and runs on macOS, Windows and Linux. It is built for work on files and code rather than personal errands, and it uses the AI subscriptions or API keys you already have."
+ - q: "Can I have more than one dot?"
+ a: "Not at launch. Each person gets one primary dot, and OpenAI says it plans to let you add more dots and have teams of dots work together later. Specialist dots with their own identity are limited to enterprise pilots."
+ - q: "How much does ChatGPT dots cost?"
+ a: "Your first dot is included with ChatGPT Pro or Business Premium at no extra cost. Pro starts at $100 a month on OpenAI's Pro tiers page, checked 30 Sep 2026. Dots are not on Free, Go or Plus."
+ - q: "Can ChatGPT dots run on my own computer?"
+ a: "The dot itself runs on OpenAI's cloud computer. You can give it optional access to your local computer through the ChatGPT desktop app, which is off by default, but the agent and its memory still live with OpenAI."
+---
+
+Pick dots for a hosted personal agent inside ChatGPT that works while your laptop is shut; pick Munder Difflin for a team of agents on your own machine using the engines you already pay for. OpenAI announced dots on 29 September 2026, and we checked everything below against its own pages on 30 September.
+
+One hosted dot, or a team on your computer.
+
+[Munder Difflin](https://harnessmd.com/download) is a free and open source desktop app. It is a [multi-agent harness](/blog/what-is-a-multi-agent-harness/): you hire agents, give each a name and a role, and they share a board on your computer. [Your first hour with Munder Difflin](/blog/your-first-hour-with-munder-difflin/) shows the setup. For the dots basics, read [what ChatGPT dots is](/blog/what-is-chatgpt-dots/).
+
+## ChatGPT dots vs Munder Difflin, checked 30 Sep 2026
+
+| | ChatGPT dots | Munder Difflin |
+| --- | --- | --- |
+| Made by | OpenAI | Open source project, MIT licence |
+| Where the agent works | OpenAI's cloud computer and browser; optional access to your computer, off by default | Terminal sessions on your own computer |
+| Agents today | One primary dot; more dots are a stated future plan | As many named agents as you hire, each with a role |
+| Models | GPT-6 Astra | 12 CLI engines, such as Claude Code, Codex and Gemini CLI, or local models via Ollama or LM Studio through OpenCode, Crush or Pi |
+| Memory | Memories from ChatGPT plus its own, including from connected apps | A plain `memory.md` file per agent on your disk |
+| Works without you asking | Proactive research (read only), scheduled tasks and reminders | Scheduled missions (interval or weekly), webhook and Slack triggers |
+| Approvals | Custom Rules per action, auto review, some tasks always stay with you | Webhook triggers held for approval by default, ASK ME board, circuit breaker |
+| Where you talk to it | ChatGPT desktop, web and mobile, Slack, Teams, voice calls, texting beta in the US | The desktop app; Slack and webhooks as inbound triggers |
+| App connections | Over 4,000 apps through plugins | MCP servers wired into each agent; write access off by default |
+| Keeps working with your laptop closed | Yes | Only if the computer running it stays on |
+| Who can get it | Pro (not EEA, Switzerland or UK) and Business Premium, age 18 and over; Enterprise, Edu and Healthcare as an admin enabled beta | Anyone; macOS, Windows and Linux |
+| Cost to start | Included in Pro or Business Premium (Pro prices below the table) | App free; you pay for your engine subscriptions or API keys |
+| Setup | Create and name it in the ChatGPT desktop app or desktop web; rollout is gradual | Install the app plus the engine CLIs you want |
+
+Dots facts come from [OpenAI's launch post](https://openai.com/index/introducing-dots/), the [getting started help article](https://help.openai.com/en/articles/20001530-getting-started-with-your-dot) and the [Pro tiers page](https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers). OpenAI's Pro tiers page listed Pro at $100, $200 or $500 a month on 30 Sep 2026. Munder Difflin facts come from the 0.5.3 release source.
+
+## Can ChatGPT dots run on my own computer?
+
+No, a dot lives on OpenAI's cloud computer with its own browser. You can open it at any time to watch or step in. If you turn on local access in the ChatGPT desktop app, the dot can also create local Work or Codex tasks, use local skills and drive your local browser. That access is off by default and you can revoke it.
+
+
+
+Munder Difflin flips this. Each agent is a terminal session on your machine, in a folder you pick. Its identity, `memory.md`, inbox and outbox are files in its workspace. [Why local first matters for AI agents](/blog/why-local-first-matters-for-ai-agents/) goes further.
+
+A dot keeps going while your laptop sleeps. A Munder Difflin office stops when the computer does.
+
+## Can I have more than one dot?
+
+Not today. At launch each person gets one primary dot. OpenAI writes that "over time, we envision teams of dots working together" and that you will be able to add more dots in the future.
+
+
+
+One dot can still work on several projects at once. In Munder Difflin you hire a team now: a reviewer, a tester, a writer, each with its own name, memory and engine. Michael, the orchestrator, reads the board, hands out tasks and keeps the team unblocked. Agents message each other through their outboxes, and the app routes each message to the right inbox. More in [running an office of AI agents](/blog/run-an-office-of-ai-agents/).
+
+## Which models can each one use?
+
+Dots run on GPT-6 Astra. Munder Difflin 0.5.3 ships presets for Claude Code, Codex, Grok, Kimi Code, Gemini CLI, Antigravity, Qwen, OpenCode, Crush, Pi, Copilot and Cursor, plus a custom slot. Local models run through OpenCode, Crush or Pi, using Ollama or LM Studio, with no API key.
+
+## How do the controls and approvals compare?
+
+Dots gate each action with Custom Rules; Munder Difflin gates inbound webhooks and throttles stuck agents. Custom Rules let you set each kind of action to act freely, act if approved in advance, ask first, or hand off to you. Auto review checks actions that touch accounts or share information, and saved passwords are used without showing them to the model.
+
+Munder Difflin's controls are aimed at a team of coding agents. A request from a webhook waits for your approval unless you loosen the mode. As of 0.5.3 the default is set in `src/shared/triggers.ts`:
+
+```ts
+export type TriggerMode = 'strict' | 'allow-all' | 'communication-only';
+// ...
+export const DEFAULT_TRIGGER_MODE: TriggerMode = 'strict';
+```
+
+When an agent needs a human decision, the card goes to the ASK ME board. A circuit breaker watches each agent for loops, repeated identical tool calls, error storms and spend over a cap you set, and escalates from a steering message to tighter limits. It only stops an agent if you turn on hard stop.
+
+## Where dots is better
+
+You can reach your dot from your phone, in Slack or Teams, or call it by voice, and over 4,000 apps plug in. Proactive research brings you things you did not ask for, using read only access.
+
+## Where Munder Difflin is better
+
+It is free and open source, with no region list. You keep the engines and subscriptions you already have. You get a team of named agents today, not a roadmap item. Memory is a markdown file you can read, edit or put in git. Files live on your disk; only what an engine reads goes to its provider. A local model keeps prompts on your machine, but agents can still fetch URLs, and anonymous usage stats stay on until you switch them off in Settings.
+
+## Is there an open source alternative to ChatGPT dots?
+
+Yes, for work on code and files: Munder Difflin is MIT licensed. It does not try to match dots on personal errands across thousands of apps. Our [ChatGPT dots alternatives](/blog/chatgpt-dots-alternatives/) list covers the others.
+
+## Should you pick ChatGPT dots or Munder Difflin?
+
+It comes down to where the work lives and how many agents you need today.
+
+- **Pick ChatGPT dots** if you have Pro outside the EEA, Switzerland and the UK, or Business Premium in any supported ChatGPT region, and want one agent for errands, inbox and research on your phone.
+- **Pick Munder Difflin** if your agents work on code or files, you want several roles now, you want to choose the model, and you want the files on your own disk. [Download it free](https://harnessmd.com/download).
diff --git a/blog/src/posts/what-is-chatgpt-dots.md b/blog/src/posts/what-is-chatgpt-dots.md
new file mode 100644
index 000000000..bea5947c2
--- /dev/null
+++ b/blog/src/posts/what-is-chatgpt-dots.md
@@ -0,0 +1,107 @@
+---
+title: "What Is ChatGPT Dots? OpenAI's Always On Agents Explained"
+seoTitle: "What Is ChatGPT Dots? Plans, Markets and Limits (Sep 2026)"
+description: "ChatGPT dots are OpenAI's always on agents, launched 29 Sep 2026. What a dot does, who can get one, what it costs and what it cannot do yet."
+date: 2026-09-30
+category: concepts
+categoryLabel: Concepts
+type: Non-technical
+primaryKeyword: "what is chatgpt dots"
+secondaryKeywords: ["chatgpt dots", "openai dots", "chatgpt dots pricing", "how to use chatgpt dots", "are chatgpt dots available in the uk"]
+tags: ["Concepts", "AI Agents", "Automation", "Human-in-the-Loop"]
+ogImage: "https://munderdiffl.in/blog/assets/media/what-is-chatgpt-dots/lead-still.png"
+faq:
+ - q: "Are ChatGPT dots available in the UK or EU?"
+ a: "Not on a personal Pro plan. At launch, Pro users in the EEA, Switzerland and the UK are excluded. Business Premium is available in all supported ChatGPT regions, so a company workspace on that plan can use dots there."
+ - q: "How much do ChatGPT dots cost?"
+ a: "Your first dot is included in ChatGPT Pro or Business Premium at no extra cost. Pro starts at $100 a month (checked 30 Sep 2026). For the first month, dots usage does not count toward plan allowances, and OpenAI says it will share usage terms after that."
+ - q: "Can I have more than one ChatGPT dot?"
+ a: "Not at launch. Each person gets one primary dot. OpenAI says you will be able to add more dots in the future and that it envisions teams of dots, but neither is available today."
+ - q: "Do ChatGPT Plus or Free users get dots?"
+ a: "No. Dots are not included in Free, Go or Plus. You need Pro, Business Premium, or an Enterprise, Edu or Healthcare workspace where an admin has turned on the beta."
+ - q: "Can a ChatGPT dot use my own computer?"
+ a: "Only if you allow it. Local computer access is off by default, and you turn it on or revoke it in the ChatGPT desktop app on that computer. Without it, the dot works on its own cloud computer and browser."
+---
+
+ChatGPT dots are always on agents that OpenAI launched on 29 September 2026 at DevDay. A dot runs on GPT-6 Astra, has its own cloud computer and browser, and takes on an ongoing responsibility rather than one task, so it keeps working between your conversations. At launch it comes with ChatGPT Pro and Business Premium.
+
+A dot works on its own cloud computer.
+
+The facts below come from [OpenAI's announcement](https://openai.com/index/introducing-dots/), its [getting started help page](https://help.openai.com/en/articles/20001530-getting-started-with-your-dot) and its [Pro tiers page](https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers), all checked on 30 Sep 2026. When we read them, the help page noted that DevDay features are rolling out gradually, and on the [ChatGPT pricing page](https://chatgpt.com/pricing) the dot was listed on the Pro card only, not on Free, Go or Plus.
+
+If what you want is a team of agents on your own computer, running the AI engines you already use, you can have that today with [Munder Difflin](https://harnessmd.com/download), free and open source. We put the two side by side in [Munder Difflin vs ChatGPT dots](/blog/munder-difflin-vs-chatgpt-dots/).
+
+Dots land three weeks after Meta Muse, and [Fortune](https://fortune.com/2026/09/29/openai-takes-on-metas-muse-with-new-dot-agents-and-unveils-a-potential-google-workspace-competitor/) framed the launch as a rival to it. We covered those hosted agents in [Meta Muse alternatives](/blog/meta-muse-alternatives/) and [Grok Bot alternatives](/blog/grok-bot-alternatives/). What sets a dot apart is that it keeps a standing job instead of a single request, the [long running agent](/blog/long-running-agents-the-2026-shift/) pattern.
+
+## What is a dot in ChatGPT?
+
+A dot is an agent that owns a piece of work over time, while a normal chat answers and then waits for you. OpenAI's own line is that dots "take on ongoing responsibility and keep making progress between conversations."
+
+Task based agents such as ChatGPT agent or a Codex cloud task take one job, run it and hand back a result. A dot can start tasks like that itself, but its unit of work is a responsibility: watch this, keep that current, chase this until it is done. [TechCrunch](https://techcrunch.com/2026/09/29/openai-launches-dots-its-bubbly-agentic-avatar/) reported OpenAI's examples: a developer whose dot monitors customer feedback and fixes the bugs it finds, and a scientist whose dot reruns an analysis when new data arrives.
+
+
+
+What a dot has to work with:
+
+* **Its own computer.** A cloud computer and browser you can open at any time to inspect its work. Access to your local computer is optional and off by default.
+* **Memory.** It receives memories from ChatGPT, creates its own (including from connected apps) and learns from your feedback.
+* **Proactive research.** When you are not working with it, it looks for ways to help using read only tools in your connected apps. In that mode it cannot send messages, change app content or control your browser or computer.
+* **Apps and Codex.** Over 4,000 apps through plugins. It can create cloud tasks in Codex, and with local access it can start Work or Codex tasks, use local skills and use your local browser.
+* **Schedules.** Scheduled tasks and reminders. Its profile shows what is In progress, Scheduled and Completed.
+
+## How do you talk to a ChatGPT dot?
+
+You message or call it in ChatGPT on desktop, web and mobile, or message it in Slack and Microsoft Teams. Texting is a limited beta for Pro users in the US, run through a third party provider. You can call your dot, but it cannot call you at launch.
+
+You name your dot during setup. The handle is `@yourname-dot` by default, or `@yourname-agentname` if you name it. The avatar is a character or a pet.
+
+## How do you control what a dot does?
+
+You set a Custom Rule per action, choosing one of four behaviours: "Take action without asking", "Take action if pre-approved", "Ask before taking action" or "Hand off to you". Built in rules already decide when to act and when to ask.
+
+
+
+On top of that, auto review checks actions that affect your accounts or share information, and some tasks, such as changing a password, always stay with you. Saved passwords are used without exposing them to the model. You can follow progress in the Activity View and pause the dot from its profile menu.
+
+Reset deletes the dot, its conversations, memories and scheduled tasks. Disconnecting an app does not delete what the dot already learned from it; only Reset does. OpenAI's warning: "Dots can still make mistakes, so always review consequential work." Our piece on [human in the loop AI agents](/blog/human-in-the-loop-ai-agents/) covers where such approvals belong.
+
+On training: Business, Enterprise and Edu content is not used by default. On personal plans you choose. OpenAI says it does not train directly on proactive research or on the dot's notes to itself.
+
+## Are ChatGPT dots available in the UK or EU?
+
+Not on a personal Pro plan. Pro users in the EEA, Switzerland and the UK are excluded at launch, while Business Premium covers all supported ChatGPT regions.
+
+**ChatGPT dots at launch, checked 30 Sep 2026**
+
+| | At launch |
+|---|---|
+| Plans | Pro 100, Pro 200, Pro 500, Business Premium. Enterprise, Edu and Healthcare get a beta an admin must turn on |
+| Not included | Free, Go, Plus |
+| Markets | Pro: outside the EEA, Switzerland and the UK. Business Premium: all supported ChatGPT regions |
+| Age | 18 and over |
+| Rollout | Gradual from 29 Sep 2026, may take several days |
+| Where to create one | ChatGPT desktop app or desktop web only |
+| Channels | ChatGPT (desktop, web, mobile), Slack, Teams, texting beta (Pro, US), voice calls to your dot |
+| Dots per person | One primary dot |
+| Own email address | No; you can connect your personal email |
+
+## How much do ChatGPT dots cost?
+
+Your first dot is included at no extra cost with Pro or Business Premium. On 30 Sep 2026 OpenAI listed Pro 100 at $100 a month, Pro 200 at $200 a month and Pro 500 at $500 a month, billed monthly only, with Astra Ultrafast only on Pro 500.
+
+Chatting with your dot does not count toward ChatGPT usage limits. Tasks it starts in Codex or ChatGPT Work do. For the first month, dots usage does not count toward eligible Pro, Business and Enterprise allowances, and OpenAI says it will share usage terms after that. OpenAI plans paid options for more dots and for more speed or work per dot, with no price yet.
+
+## What can't ChatGPT dots do yet?
+
+At launch a dot cannot:
+
+* Be one of several. You get one primary dot. "Over time, we envision teams of dots working together," OpenAI writes, and specialist dots with their own identity and credentials are in enterprise pilots only.
+* Have its own email address.
+* Call you. You call it.
+* Be created on mobile or mobile web.
+
+## Who should use ChatGPT dots?
+
+If you already pay for Pro outside the EEA, Switzerland and the UK, set one up: the first month's usage is not counted against your plan. Dots suit personal and business errands across many apps, with zero setup, a mobile app, voice, and a hosted computer that works while your laptop is closed.
+
+They suit you less if you are in the UK or EU on a personal plan, if you want several agents with separate roles today, or if you want agents on your own machine running Claude Code, Codex and Gemini CLI side by side with your files kept local. For those cases, see [ChatGPT dots alternatives](/blog/chatgpt-dots-alternatives/).
diff --git a/docs/blog/agents-rule-of-two/index.html b/docs/blog/agents-rule-of-two/index.html
index c9345494b..863149b97 100644
--- a/docs/blog/agents-rule-of-two/index.html
+++ b/docs/blog/agents-rule-of-two/index.html
@@ -331,50 +331,50 @@
OpenCode is Anomaly's MIT licensed coding agent for the terminal. What it is, how to install it, what it costs, and how it compares with Claude Code and Codex.
OpenCode is Anomaly's MIT licensed coding agent for the terminal. What it is, how to install it, what it costs, and how it compares with Claude Code and Codex.
OpenCode is Anomaly's MIT licensed coding agent for the terminal. What it is, how to install it, what it costs, and how it compares with Claude Code and Codex.
OpenCode is Anomaly's MIT licensed coding agent for the terminal. What it is, how to install it, what it costs, and how it compares with Claude Code and Codex.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.
Personal ChatGPT dots need Pro from $100 a month and skip the UK and EU. Seven alternatives, sorted by where they run: your computer or a vendor's cloud.
Meta Muse is US only and trains on your chats by default. Seven alternatives, from open source agents that run on your own computer to hosted ones from OpenAI, Google and SpaceXAI.
Meta Muse and Grok Bot both ask for your logins. We read both companies' security docs: who sees your passwords, what trains on your chats, and how to keep AI agents on your own computer instead.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.
Personal ChatGPT dots need Pro from $100 a month and skip the UK and EU. Seven alternatives, sorted by where they run: your computer or a vendor's cloud.
Meta Muse is US only and trains on your chats by default. Seven alternatives, from open source agents that run on your own computer to hosted ones from OpenAI, Google and SpaceXAI.
Meta Muse and Grok Bot both ask for your logins. We read both companies' security docs: who sees your passwords, what trains on your chats, and how to keep AI agents on your own computer instead.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.
Personal ChatGPT dots need Pro from $100 a month and skip the UK and EU. Seven alternatives, sorted by where they run: your computer or a vendor's cloud.
Meta Muse is US only and trains on your chats by default. Seven alternatives, from open source agents that run on your own computer to hosted ones from OpenAI, Google and SpaceXAI.
Meta Muse and Grok Bot both ask for your logins. We read both companies' security docs: who sees your passwords, what trains on your chats, and how to keep AI agents on your own computer instead.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.
OpenCode is Anomaly's MIT licensed coding agent for the terminal. What it is, how to install it, what it costs, and how it compares with Claude Code and Codex.
OpenCode is Anomaly's MIT licensed coding agent for the terminal. What it is, how to install it, what it costs, and how it compares with Claude Code and Codex.
ChatGPT Dots Alternatives: 7 Always On Agents, Local and Cloud
+
Personal ChatGPT dots need Pro from $100 a month and skip the UK and EU. Seven alternatives, sorted by where they run: your computer or a vendor's cloud.
The best ChatGPT dots alternatives are OpenClaw, Munder Difflin and Hermes Agent if you want agents on your own computer, and Claude Cowork, Gemini Spark, Grok Bot or Meta Muse if you want another hosted agent.
+
TL;DR
A free personal assistant you own: OpenClaw. A team of work agents on your computer: Munder Difflin. Memory that learns and a cron you control: Hermes Agent. Work that keeps going with your laptop shut: Claude Cowork or Grok Bot. Life in Google apps: Gemini Spark. A free hosted agent in the US or Canada: Meta Muse.
This settles who holds your files and whether work stops when you close the lid. Three picks run on hardware you control, four on a vendor’s cloud, like dots. We drew the same line in our Grok Bot alternatives guide, and why local first matters for AI agents explains which side we lean to.
1. OpenClaw.OpenClaw is an open source (MIT) personal assistant that runs on your own computer and answers in WhatsApp, Telegram, Discord, Slack, Teams, iMessage and 20+ more channels. It runs as a background daemon, keeps memory and credentials on your hardware, and has no paid tier. Best for: a personal errand runner you own.
+
2. Munder Difflin.Munder Difflin is what we make: a free and open source desktop app that runs a team of named AI agents on your own computer. Each agent has a role and runs a CLI engine you already use: Claude Code, Codex, Gemini CLI, OpenCode, Cursor and more. Each one keeps its own memory.md, inbox and outbox. You can put missions on a schedule, and a Slack channel can feed work straight to Michael, your lead agent. The catch: agents run while your computer runs, and you install the CLIs yourself. Best for: engineering and project work across several engines. (Munder Difflin 0.5.3, checked 30 Sep 2026.)
+
3. Hermes Agent.Hermes Agent from Nous Research is open source (MIT) and built around a learning loop: it curates its own memory and writes new skills from experience. Its cron scheduler runs jobs unattended, and you reach it over Telegram, Discord, Slack, WhatsApp or Signal. Best for: memory that grows, on a box you control.
4. Claude Cowork. Cowork is Anthropic’s agent for longer pieces of work, and it now runs in the cloud. Anthropic’s Cowork help page says “Claude’s work runs on Anthropic’s servers, in an isolated environment” and that scheduled tasks don’t need your computer to be awake. Since 16 Sep 2026 Anthropic has been folding Cowork into the main Claude app, Pro and Max first. Claude Pro was $20 a month, billed monthly, on 30 Sep 2026. Best for: reports, files and research, task by task.
+
5. Gemini Spark. Spark is Google’s personal agent in the Gemini app. It works from your Connected Apps and chats, and can drive your local Chrome. If your device goes off before a task is done, Google’s Spark help page says it “might use a remote browser” in the cloud instead. You need Google AI Pro or Ultra and to be 18 or over. Like dots on Pro, it is not available in the EEA, Switzerland or the UK, nor in Nigeria. Best for: Google app users outside Europe.
+
6. Grok Bot. SpaceXAI’s docs say your Bots run on “a persistent cloud computer with a browser, filesystem, and terminal”, and closing your laptop or phone does not stop a routine. It comes with paid Cursor plans or a SuperGrok account. Cursor’s Individual plan was $20 a month on 30 Sep 2026. Grok Bot pricing has the other plans. Best for: hosted work agents whose computer never sleeps.
+
7. Meta Muse.Meta’s help centre says Muse runs on “a cloud-based environment called a virtual machine”. On 29 Sep 2026 Meta said Muse is “available in the US and Canada” and free for most use; its subscriptions page lists Power at $20 a month and Max at $100. Best for: a free hosted agent in North America. Elsewhere, see Meta Muse alternatives.
Yes: OpenClaw, Munder Difflin and Hermes Agent are all free and open source. The software costs nothing, but the model still does. OpenClaw and Hermes Agent take API keys, local models or a provider login. Munder Difflin drives the CLIs you already have, so the Claude or ChatGPT subscription you pay for today does the work.
+
+
Meta Muse is the only hosted pick with a free tier, and only in the US and Canada. Dots start at $100 a month on Pro (30 Sep 2026), and OpenAI has not said how dots usage will count against your plan after the first month.
+
What is the best open source alternative to ChatGPT dots? #
+
OpenClaw, if you want what a dot does for your personal life: one assistant, in your chat apps, on your hardware. Munder Difflin, if you want agents for work: several named agents with roles, each on the engine that suits the job. Hermes Agent is strongest on memory.
+
None of the three is a hosted computer that works with your laptop shut: you need a server, or a machine that stays on. We put Munder Difflin and dots side by side in Munder Difflin vs ChatGPT dots.
+
+
+
+
+
FAQ
+
+
+
Is there a free alternative to ChatGPT dots?
+
Yes. OpenClaw, Munder Difflin and Hermes Agent are free and open source and run on hardware you control. You still pay for the AI model, or use one you already subscribe to. Meta Muse has a free hosted tier, but only in the US and Canada as of 29 Sep 2026.
+
+
+
+
Can I use ChatGPT dots in the UK or EU?
+
Not on a personal Pro plan. At launch OpenAI excludes Pro users in the EEA, Switzerland and the UK, while Business Premium covers all supported ChatGPT regions. Gemini Spark has the same gap, so the open source picks are the simplest route there.
+
+
+
+
What is the best open source alternative to ChatGPT dots?
+
OpenClaw for a personal assistant in your chat apps, Munder Difflin for a team of work agents on your computer, and Hermes Agent for memory that learns over time. All three are open source and run on your own machine or server.
+
+
+
+
Which ChatGPT dots alternatives keep working when my laptop is off?
+
The cloud ones. Claude Cowork runs scheduled tasks on Anthropic's servers, Grok Bot works on a persistent cloud computer, and Gemini Spark might finish a task in a remote browser. The local picks need a machine that stays on.
+
+
+
+
Can I run more than one agent instead of a single dot?
+
Not with dots yet: at launch each person gets one primary dot. Munder Difflin runs several named agents with separate roles on your computer today, each on the CLI engine you pick.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.
Personal ChatGPT dots need Pro from $100 a month and skip the UK and EU. Seven alternatives, sorted by where they run: your computer or a vendor's cloud.
Meta Muse is US only and trains on your chats by default. Seven alternatives, from open source agents that run on your own computer to hosted ones from OpenAI, Google and SpaceXAI.
Meta Muse and Grok Bot both ask for your logins. We read both companies' security docs: who sees your passwords, what trains on your chats, and how to keep AI agents on your own computer instead.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.
Personal ChatGPT dots need Pro from $100 a month and skip the UK and EU. Seven alternatives, sorted by where they run: your computer or a vendor's cloud.
Meta Muse is US only and trains on your chats by default. Seven alternatives, from open source agents that run on your own computer to hosted ones from OpenAI, Google and SpaceXAI.
Meta Muse and Grok Bot both ask for your logins. We read both companies' security docs: who sees your passwords, what trains on your chats, and how to keep AI agents on your own computer instead.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.
Personal ChatGPT dots need Pro from $100 a month and skip the UK and EU. Seven alternatives, sorted by where they run: your computer or a vendor's cloud.
Meta Muse is US only and trains on your chats by default. Seven alternatives, from open source agents that run on your own computer to hosted ones from OpenAI, Google and SpaceXAI.
Meta Muse and Grok Bot both ask for your logins. We read both companies' security docs: who sees your passwords, what trains on your chats, and how to keep AI agents on your own computer instead.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.
Personal ChatGPT dots need Pro from $100 a month and skip the UK and EU. Seven alternatives, sorted by where they run: your computer or a vendor's cloud.
Meta Muse is US only and trains on your chats by default. Seven alternatives, from open source agents that run on your own computer to hosted ones from OpenAI, Google and SpaceXAI.
Meta Muse and Grok Bot both ask for your logins. We read both companies' security docs: who sees your passwords, what trains on your chats, and how to keep AI agents on your own computer instead.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.
Personal ChatGPT dots need Pro from $100 a month and skip the UK and EU. Seven alternatives, sorted by where they run: your computer or a vendor's cloud.
Meta Muse is US only and trains on your chats by default. Seven alternatives, from open source agents that run on your own computer to hosted ones from OpenAI, Google and SpaceXAI.
Meta Muse and Grok Bot both ask for your logins. We read both companies' security docs: who sees your passwords, what trains on your chats, and how to keep AI agents on your own computer instead.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.
Personal ChatGPT dots need Pro from $100 a month and skip the UK and EU. Seven alternatives, sorted by where they run: your computer or a vendor's cloud.
Meta Muse is US only and trains on your chats by default. Seven alternatives, from open source agents that run on your own computer to hosted ones from OpenAI, Google and SpaceXAI.
Meta Muse and Grok Bot both ask for your logins. We read both companies' security docs: who sees your passwords, what trains on your chats, and how to keep AI agents on your own computer instead.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.
Personal ChatGPT dots need Pro from $100 a month and skip the UK and EU. Seven alternatives, sorted by where they run: your computer or a vendor's cloud.
Meta Muse is US only and trains on your chats by default. Seven alternatives, from open source agents that run on your own computer to hosted ones from OpenAI, Google and SpaceXAI.
Meta Muse and Grok Bot both ask for your logins. We read both companies' security docs: who sees your passwords, what trains on your chats, and how to keep AI agents on your own computer instead.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.
Personal ChatGPT dots need Pro from $100 a month and skip the UK and EU. Seven alternatives, sorted by where they run: your computer or a vendor's cloud.
Meta Muse is US only and trains on your chats by default. Seven alternatives, from open source agents that run on your own computer to hosted ones from OpenAI, Google and SpaceXAI.
Meta Muse and Grok Bot both ask for your logins. We read both companies' security docs: who sees your passwords, what trains on your chats, and how to keep AI agents on your own computer instead.
Grok Bot has no plan of its own. It comes bundled with SuperGrok and Cursor plans from $20 to $300 a month. Here is what each one gets you, what the free trial is, and the $0 route for work agents.