Skip to content

add screenpipe solution profile (end-to-end apps) #25

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
louis030195 wants to merge 1 commit into
base: main
Choose a base branch
from
Open

add screenpipe solution profile (end-to-end apps) #25

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,6 +40,7 @@ Start with the second-brain lifecycle stage that is blocking you most. If you wa
|---|---|---|
| Collect scattered context | [Membase](solutions/membase.md), [OpenHuman](solutions/openhuman.md), [Supermemory](solutions/supermemory.md), [Hyperspell](solutions/hyperspell.md), [Khoj](solutions/khoj.md), or [Obsidian/Logseq + AI bridge](solutions/obsidian-logseq.md) | Use these when chats, docs, notes, files, apps, and workspace sources are not yet flowing into a usable brain. |
| Organize raw context into durable knowledge | [Membase](solutions/membase.md), [GBrain](solutions/gbrain.md), [Hermes Agent + LLM Wiki](solutions/hermes-llm-wiki.md), [Hermes Agent + Obsidian + Honcho](solutions/hermes-obsidian-honcho.md), [Mnemosyne](solutions/mnemosyne.md), [taOSmd](solutions/taosmd.md), [Honcho](solutions/honcho.md), [Hjarni](solutions/hjarni.md), [Zep/Graphiti](solutions/zep-graphiti.md), [Cognee](solutions/cognee.md), or [Pad](solutions/pad.md) | Use these when raw context needs memory records, wiki pages, facts, links, graphs, timelines, or other durable structure. |
| [screenpipe](solutions/screenpipe.md) | Collect, Use | You want fully automatic, local-first capture of everything on your screen and in meetings, queryable by agents over MCP, API, or CLI. | Always-on capture is Collect-heavy, so you own retention and redaction, and durable organization is built on top via pipes or agents. |
| Evolve memory over time | [Membase](solutions/membase.md), [GBrain](solutions/gbrain.md), [Hermes Agent + Obsidian + Honcho](solutions/hermes-obsidian-honcho.md), [Hyperspell](solutions/hyperspell.md), [Honcho](solutions/honcho.md), [Hindsight](solutions/hindsight.md), [Mnemosyne](solutions/mnemosyne.md), [taOSmd](solutions/taosmd.md), [Zep/Graphiti](solutions/zep-graphiti.md), or [Cognee](solutions/cognee.md) | Use these when new context should update, consolidate, dedupe, refresh, or re-reason over existing memory. |
| Use context inside AI tools and workflows | [Membase](solutions/membase.md), [Supermemory](solutions/supermemory.md), [Hyperspell](solutions/hyperspell.md), [Honcho](solutions/honcho.md), [Hindsight](solutions/hindsight.md), [Mnemosyne](solutions/mnemosyne.md), [taOSmd](solutions/taosmd.md), [Hjarni](solutions/hjarni.md), [Mem0/OpenMemory](solutions/mem0-openmemory.md), [Claude Projects/Claude Code](solutions/claude-projects-code.md), or [Pad](solutions/pad.md) | Use these when the main need is MCP, API, SDK, plugin, dashboard chat, or platform access that puts memory into active work. |
| Govern, inspect, correct, or control memory | [Membase](solutions/membase.md), [GBrain](solutions/gbrain.md), [taOSmd](solutions/taosmd.md), [Hermes Agent + LLM Wiki](solutions/hermes-llm-wiki.md), [Obsidian/Logseq + AI bridge](solutions/obsidian-logseq.md), [Hjarni](solutions/hjarni.md), [ChatGPT Memory](solutions/chatgpt-memory.md), [Claude Projects/Claude Code](solutions/claude-projects-code.md), or [Pad](solutions/pad.md) | Use these when visibility, review, correction, deletion, ownership, permissions, or local/cloud control matter most. |
Expand Down
77 changes: 77 additions & 0 deletions solutions/screenpipe.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,77 @@
# screenpipe

## Snapshot

- Website / docs: https://screenpi.pe and https://docs.screenpi.pe
- Repo: https://github.com/screenpipe/screenpipe
- Company / maintainer: screenpipe (YC S26)
- Status: Production, actively developed with frequent releases
- Open source: Source-available; the CLI is free and self-hostable
- Deployment: Local-first desktop app and CLI daemon on macOS, Windows, and Linux; capture, OCR, transcription, indexing, and storage run on-device in SQLite under `~/.screenpipe`. Cloud AI and cloud transcription are optional and opt-in.
- Primary users: People who want continuous, automatic local capture of their screen and audio as searchable memory for themselves and their AI agents
- Best second-brain role: Local-first Collect substrate that feeds agents real context
- Last reviewed: 2026-06-16

## One-line Summary

screenpipe continuously records your screen (accessibility tree with OCR fallback) and audio (transcription with speaker diarization), indexes everything in a local SQLite database, and exposes it to AI agents through an MCP server, a local REST API, and a CLI.

## Second-Brain Fit

screenpipe is best evaluated as a Collect-layer capture substrate rather than a governed knowledge base. It automatically turns everything you see, say, and hear into on-device, searchable memory that agents can query. Its strength is broad, effortless capture and local ownership; turning that raw stream into durable, organized knowledge is something you build on top via "pipes" (scheduled automations) or your own agents, not a built-in knowledge graph.

## Capabilities

| Area | Evaluation |
|---|---|
| Deployment / ownership | Desktop app and CLI daemon on macOS, Windows, and Linux; capture, OCR, transcription, indexing, and SQLite storage all on-device under `~/.screenpipe`. Cloud AI/transcription optional and opt-in. |
| Context capture | Automatic 24/7 screen capture (accessibility-first text with OCR fallback), microphone and system audio (transcription with speaker diarization), input events (clicks, keystrokes, clipboard), and meeting detection. |
| Knowledge organization | Timestamped events in local SQLite with full-text and embedding search. Pipes can segment and summarize activity into Markdown. No built-in governed knowledge graph; organization is assembled via pipes or agents. |
| Memory evolution | Not automatic by default; consolidation and summarization are done through pipes or external agents you configure. |
| Retrieval / use | Local REST API (`localhost:3030`), MCP server (`npx -y screenpipe-mcp`, stdio or HTTP), and CLI. Search by text, time, app, window, or speaker; activity summaries, meeting lists, and video export. |
| Agent activation / write-back | First-class MCP server gives agents (Claude, OpenClaw, Hermes, and others) read access to your history; write-back is via pipes and the `update-memory` tool. |
| Personal / team scope | Personal-first. An enterprise tier adds MDM deployment and admin features for organizations. |
| Feedback / correction | Privacy controls include app/domain ignore lists, recording schedules, PII redaction, and default-off options. Data is fully deletable (`rm -rf ~/.screenpipe`). |
| Privacy / control | Local-first by design; data stays on device unless you opt into cloud features. Full inspect, export, and delete. |
| Setup / operations | Low-medium. One-click app install or `brew install screenpipe`; querying via MCP/API/CLI is straightforward; building organized knowledge requires configuring pipes or agents. |

## Strengths

- Broadest automatic capture of the tools compared here: screen (OCR plus accessibility) and audio, 24/7, not just notes or app connectors.
- Fully local-first: capture, OCR, transcription, and storage on-device by default, with cloud as an explicit opt-in.
- Strong agent activation: official MCP server plus a REST API and a CLI.
- Cross-platform (macOS, Windows, Linux) and scriptable as a CLI daemon.

## Limitations

- Collect-heavy: it excels at capture and retrieval, but durable Organize/Evolve/Govern (knowledge graph, automatic consolidation) is assembled on top via pipes or agents.
- Always-on capture means you own retention, redaction, and privacy decisions.
- Source-available under a commercial license rather than an OSI license; the CLI is free and self-hostable.

## Best For

- Users who want automatic, no-effort capture of everything they see and hear as local, searchable memory.
- People feeding a personal agent (OpenClaw, Hermes, Claude) real context about their day.
- Local, on-device replacements for Rewind, Limitless, or Microsoft Recall.

## Not Ideal For

- Users who want a built-in, governed knowledge graph with automatic consolidation out of the box.
- Teams needing shared workspace memory governance with no additional setup (enterprise tier aside).
- Contexts where always-on screen capture is not acceptable.

## Tradeoffs

screenpipe maximizes the Collect layer: it captures more raw personal context, locally, than note- or connector-based tools, and exposes it to agents over MCP, API, and CLI. The tradeoff is that turning that raw stream into organized, governed knowledge is something you assemble (via pipes or agents), and always-on capture puts retention and redaction decisions on you.

## Official Setup / Evaluation Links

- [screenpipe docs](https://docs.screenpi.pe)
- [OpenClaw + screenpipe integration guide](https://docs.screenpi.pe/openclaw)
- [screenpipe-mcp package](https://github.com/screenpipe/screenpipe/tree/main/packages/screenpipe-mcp)

## Sources

- [screenpipe GitHub repo](https://github.com/screenpipe/screenpipe)
- [screenpipe docs](https://docs.screenpi.pe)
- [screenpipe-mcp package](https://github.com/screenpipe/screenpipe/tree/main/packages/screenpipe-mcp)