MCPFind · gusmar2017 · Aug 3, 2026 · Jul 29, 2026
diff --git a/apps/web/content/blog/claude-skills-vs-mcp-servers.mdx b/apps/web/content/blog/claude-skills-vs-mcp-servers.mdx
@@ -0,0 +1,94 @@
+---
+title: "Claude Skills vs MCP Servers: What's the Difference?"
+description: "Claude Skills and MCP servers solve different problems in the same agent stack. Here is how each works, what they cost in context, and when to build which one."
+excerpt: "Claude Skills and MCP servers are not competitors. Skills package instructions an agent loads on demand; MCP servers connect that agent to systems outside its own process. This guide breaks down how each works, what each costs in context tokens, and how to decide which one to build next."
+date: "2026-07-21"
+updatedAt: "2026-07-21"
+author: "Gus Marquez"
+authorUrl: "https://www.linkedin.com/in/gustavoamarquez"
+authorSameAs:
+  - "https://www.linkedin.com/in/gustavoamarquez"
+tags:
+  - "mcp"
+  - "developer"
+  - "architecture"
+  - "claude-skills"
+  - "how-to"
+category: "other"
+focusKeyword: "claude skills vs mcp servers"
+draft: false
+cornerstone: false
+topic_source: new_topic
+faqItems:
+  - question: "Can Claude use Skills and MCP servers in the same conversation?"
+    answer: "Yes. A Skill can call tools exposed by an MCP server as part of its instructions, and an agent can have both loaded at once. They are not mutually exclusive configurations."
+  - question: "Do Skills replace the need for MCP servers?"
+    answer: "No. A Skill without a connected data source or API still cannot read your database, post to Slack, or query a live system. MCP servers are what give Skills (and any other agent) that external reach."
+  - question: "Why did MCP move to the Agentic AI Foundation instead of staying under Anthropic?"
+    answer: "Multi-vendor governance was the point. Once Google, Microsoft, AWS, Cloudflare, and Bloomberg were all shipping MCP integrations, a foundation structure signals the protocol will not bend to any single company's roadmap, which matters if you are building on it long term."
+  - question: "Is there a directory for Claude Skills the way MCPFind is a directory for MCP servers?"
+    answer: "Not yet at the same scale. Skills are newer and mostly distributed through individual repos and Anthropic's own examples, while MCP has years of registry infrastructure (MCPFind included) already built around it."
+---
+
+Claude Skills and MCP servers hit the news cycle close enough together that people keep asking which one wins, but the versus framing falls apart the moment you look at what each actually does. A Skill is a folder of instructions, scripts, and reference files that Claude loads only when a task calls for it. An MCP server is a separate process. It exposes tools, data, or prompts over a standard protocol, and any MCP client can reach it, whether that's Claude Desktop, Cursor, or something you wrote yourself. We currently index 18,209 MCP servers in the MCPFind directory, and a Skill can't replace a single one of them, because a Skill still needs something to connect to.
+
+## What Is the Difference Between Claude Skills and MCP Servers?
+
+A Claude Skill is a bundle of markdown instructions, plus the occasional helper script, that Claude pulls into context only when a task matches what the Skill covers. An MCP server is different in kind. It's a running process, local or remote, exposing tools over stdio, SSE, or HTTP so any compatible client can call them. A PDF-generation Skill spells out how Claude should format an invoice; a Postgres MCP server is what actually pulls the line items that go into it.
+
+Anthropic frames it the same way. Claude ships with "75+ connectors powered by MCP" right now, and Skills sit on top of that connector layer instead of replacing it. Think about a Skill for writing quarterly reports. It still needs an MCP server, or a built-in tool, to actually pull the sales numbers it's reporting on. Cut off the data source and the Skill is well-organized advice with nothing to act on. Get that distinction straight first. Everything else in this comparison depends on it.
+
+## How Do Claude Skills and MCP Servers Work Together?
+
+Skills orchestrate and MCP servers connect. A Skill's instructions can name specific MCP tools and tell Claude exactly when and how to call them inside a multi-step task, while the server does the unglamorous work of actually reaching your database, ticketing system, or file store.
+
+Picture a "close the sprint" Skill. The instructions say: check open pull requests, flag the ones without reviews, post a summary to the team channel. Not one of those steps runs without a GitHub MCP server and a Slack MCP server already wired up.
+
+```json
+{
+  "mcpServers": {
+    "github": { "command": "npx", "args": ["-y", "@modelcontextprotocol/server-github"] },
+    "slack": { "command": "npx", "args": ["-y", "@modelcontextprotocol/server-slack"] }
+  }
+}
+```
+
+With both servers configured, the Skill references their tools by name in its instructions and Claude resolves the calls on its own. MCPFind's [devtools category](/categories/devtools) alone lists 5,082 servers a Skill like this could draw from, everything from repo intelligence to CI log parsing.
+
+There's a knock-on effect here. Teams already running several MCP servers tend to add Skills one at a time rather than in a big push, and the reason is cost. The hard part, the actual connections, is already done. Writing the workflow instructions on top of it takes an afternoon, not a new deployment.
+
+## Why Was MCP Donated to the Agentic AI Foundation?
+
+MCP moved out from under Anthropic and into a directed fund at the Linux Foundation called the Agentic AI Foundation. The transfer finalized on December 9, 2025. What it did was hand day-to-day stewardship of the protocol to a multi-vendor foundation. Anthropic, Block, and OpenAI co-founded the foundation, with Google, Microsoft, AWS, Cloudflare, and Bloomberg backing it. MCP joined as a founding project next to Block's goose agent and OpenAI's AGENTS.md spec.
+
+The real driver is scale. Anthropic's announcement noted MCP had crossed 10,000 active public servers inside its first year, with every major agent vendor shipping integrations against it. When a protocol carries that much cross-vendor weight, it needs neutral stewardship. Otherwise every roadmap call looks like it favors whoever owns the repo. Foundation governance is the standard fix for that exact problem, and it's the same route AGENTS.md and goose took by joining the same body. If you're building against MCP for the long haul, read this as a stability signal. The protocol itself works the same day to day.
+
+## How Much Context Do Skills Save Compared to Loading MCP Tools?
+
+Skills are dramatically cheaper on context, right up until you actually use them. A Skill's metadata runs roughly 30 to 50 tokens sitting idle in Claude's context window, because only the name and a one-line description load by default. The full instructions expand into context later, once Claude decides the task matches.
+
+MCP tools don't behave that way. Connect five servers, each with a typical tool count, and you can blow past 50,000 tokens of tool schemas before a single tool ever gets called, because every connected server dumps its full tool list up front. That gap is exactly why teams increasingly wrap heavily-used MCP tool chains inside a Skill. The Skill's thin metadata stays in context by default. The expensive tool schemas only load when the Skill decides they're needed. Once your tool count grows past a handful of servers, this progressive-disclosure pattern becomes the default way serious agent builders keep their context budget under control.
+
+## When Should You Build a Skill Instead of an MCP Server?
+
+Build a Skill when you're packaging a repeatable process Claude can already carry out. Report templates. Code review checklists. Multi-step runbooks. All of that is Skill territory, because Claude already has, or can already reach, everything it needs to run them. What's missing is the instructions for doing it your team's specific way.
+
+Build an MCP server when Claude flat-out can't reach something yet. If no existing tool queries your internal API, reads your specific database, or posts to your specific ticketing system, you're looking at a server problem, not an instructions problem. No amount of well-written Skill markdown stands in for a real connection. And most real projects end up needing both: an MCP server (or several, pulled from [MCPFind's ai-ml category](/categories/ai-ml) and its 2,100 servers) to establish reach, then a Skill on top to make that reach usable. New to servers? The [plain-English MCP guide](/blog/what-is-mcp) starts from zero, and the [Go MCP server tutorial](/blog/build-mcp-server-go-tutorial) walks you through building one once you've spotted the gap you're filling.
+
+## Frequently Asked Questions
+
+### Can Claude use Skills and MCP servers in the same conversation?
+
+Yes. A Skill can call tools exposed by an MCP server as part of its instructions, and an agent can have both loaded at once. They are not mutually exclusive configurations.
+
+### Do Skills replace the need for MCP servers?
+
+No. A Skill without a connected data source or API still cannot read your database, post to Slack, or query a live system. MCP servers are what give Skills (and any other agent) that external reach.
+
+### Why did MCP move to the Agentic AI Foundation instead of staying under Anthropic?
+
+Multi-vendor governance was the point. Once Google, Microsoft, AWS, Cloudflare, and Bloomberg were all shipping MCP integrations, a foundation structure signals the protocol will not bend to any single company's roadmap, which matters if you are building on it long term.
+
+### Is there a directory for Claude Skills the way MCPFind is a directory for MCP servers?
+
+Not yet at the same scale. Skills are newer and mostly distributed through individual repos and Anthropic's own examples, while MCP has years of registry infrastructure (MCPFind included) already built around it.
diff --git a/apps/web/content/blog/elevenlabs-mcp-server-voice-ai.mdx b/apps/web/content/blog/elevenlabs-mcp-server-voice-ai.mdx
@@ -0,0 +1,101 @@
+---
+title: "ElevenLabs MCP Server: Voice Cloning and TTS via Claude"
+description: "Set up the ElevenLabs MCP server to add text-to-speech and voice cloning to Claude. Covers install, tools exposed, and how it compares to alternatives."
+excerpt: "The ElevenLabs MCP server puts text-to-speech, voice cloning, and transcription directly into Claude's toolset. This guide covers what the server exposes, how to install it, and where it fits next to MCPFind's media and ai-ml category servers."
+date: "2026-07-22"
+updatedAt: "2026-07-22"
+author: "Gus Marquez"
+authorUrl: "https://www.linkedin.com/in/gustavoamarquez"
+authorSameAs:
+  - "https://www.linkedin.com/in/gustavoamarquez"
+tags:
+  - "mcp"
+  - "developer"
+  - "ai-ml"
+  - "voice-ai"
+  - "elevenlabs"
+category: "media"
+focusKeyword: "elevenlabs mcp server"
+draft: false
+cornerstone: false
+topic_source: new_topic
+faqItems:
+  - question: "Does the ElevenLabs MCP server require a paid account?"
+    answer: "You need an ElevenLabs API key, and voice cloning and higher usage tiers require a paid plan. The free tier covers basic text-to-speech testing but hits character limits quickly for real workloads."
+  - question: "Can the ElevenLabs MCP server run alongside other audio tools in the same agent?"
+    answer: "Yes. It exposes its own named tools, so it does not conflict with a transcription-only server or a different TTS provider connected in parallel. Claude picks the right tool based on what the task calls for."
+  - question: "Does the ElevenLabs MCP server work with Cursor and Windsurf, or only Claude?"
+    answer: "It works with any MCP-compatible client, including Cursor, Windsurf, and the OpenAI Agents SDK, since the server speaks the standard protocol rather than a Claude-specific extension."
+  - question: "Is voice cloning through the MCP server subject to ElevenLabs' consent requirements?"
+    answer: "Yes. ElevenLabs enforces its own voice cloning consent and usage policies at the API level, and the MCP server does not bypass them. You still need explicit rights to any voice you clone."
+---
+
+Most MCP servers hook Claude up to data. The ElevenLabs MCP server hooks it up to a voice. Wire this one into Claude Desktop, Cursor, or any other MCP client, and text-to-speech, voice cloning, transcription, and a handful of audio-editing tools all go live at once. Compare that to the database and API integrations that fill out most of MCPFind's [media category](/categories/media), and you're looking at a different animal entirely. It cracks open workflows plain text can't reach: agent narration, spoken reports, voice-cloned assistants, automated dubbing.
+
+Below, we cover what the server actually exposes, how to install it, where the community alternatives come in when the official option falls short, and the real-world limits worth knowing before you build on top of it.
+
+## What Does the ElevenLabs MCP Server Let Claude Do?
+
+Text-to-speech, voice cloning, speech-to-text transcription, voice design, audio isolation, soundscape generation. The ElevenLabs MCP server hands all of those to Claude as callable tools. So Claude can narrate a written report, clone a reference voice off a short sample, transcribe a recorded meeting, or scrub background noise out of an existing clip, and it does all of that inside the same conversation that generated the text in the first place.
+
+Here is the split that actually matters: generation versus manipulation. Voice cloning and voice design build new audio from a description or a sample. Transcription and audio isolation work on audio you already have. Most workflows land on one side of that line, not both. That makes your first setup decision an easy one to name and an easy one to get wrong: which tools do you genuinely need turned on, versus which ones just pile up surface area you now have to review? Enabling everything by default is rarely right on a first pass. With more tools active, Claude has a larger menu to choose from, and the odds of a wrong pick go up accordingly.
+
+## How Do You Install and Configure the ElevenLabs MCP Server?
+
+Installation is the standard MCP dance. Drop the server into your client's config with an API key, then restart the client so it picks up the new tools. Depending on which distribution you pull, the server runs through `uvx` or `npx`, and it needs nothing more than a valid ElevenLabs API key to start doing work.
+
+```json
+{
+  "mcpServers": {
+    "elevenlabs": {
+      "command": "uvx",
+      "args": ["elevenlabs-mcp"],
+      "env": {
+        "ELEVENLABS_API_KEY": "your_api_key_here"
+      }
+    }
+  }
+}
+```
+
+After it connects, Claude Desktop or Cursor lists the new tools on its own. No separate registration step. Start with something dead simple like "read this paragraph back to me" before you touch voice cloning. Cloning burns API credits faster, so you want the basics confirmed before you start spending on it.
+
+One catch before you commit to an implementation. MCPFind's directory currently indexes three separate ElevenLabs MCP wrappers, each under a different maintainer, and none of them shows up as an official vendor listing in our registry data. For a fast-moving integration space that is normal enough. But it does mean the config snippet above might need a tweak depending on which package name you end up pulling. Look at the tool list a package exposes before you wire it into anything production-bound, because coverage of voice cloning, transcription, and audio isolation shifts from one implementation to the next even when they all wrap the same underlying API.
+
+## What Are the Tool-Level Limits Worth Knowing Before You Rely on It?
+
+Run this past quick testing and two things surface fast: character limits and voice cloning consent. Free-tier accounts blow through a low monthly character cap in a hurry, so anything beyond a test drive needs a paid plan. Voice cloning carries its own rule. You have to hold rights to the voice you are cloning, and ElevenLabs enforces that at the API level, so the MCP server just passes your request straight through. Confirm you hold those rights before you start building on top of it.
+
+Latency is the other one. Generating a few minutes of audio takes noticeably longer than a text response, so any agent chaining TTS calls back to back should expect the conversation to stall on each generation instead of streaming instantly. None of this is a quirk of the MCP wrapper. It's just how the underlying ElevenLabs API behaves. These are API-level constraints, and they show up the same way no matter how you reach the service.
+
+## How Does the ElevenLabs MCP Server Compare to Other Voice Options in the Directory?
+
+Voice generation is a thin slice of what's in there. MCPFind's [media category](/categories/media) currently indexes 517 servers, most of them leaning toward video, images, or file conversion rather than speech specifically. A few community-maintained ElevenLabs wrappers sit inside that set, right alongside the vendor's own tooling, and each one packages a slightly different slice of the underlying API. So check what a given listing actually exposes. Don't assume full API parity.
+
+Got a narrower need? Local transcription only, say. Then a dedicated speech-to-text server, minus the whole generation and cloning surface, probably beats pulling in the full feature set. If voice is just one piece of a bigger pipeline, the broader [ai-ml category](/categories/ai-ml) and its 2,100 indexed servers is where most of the surrounding model-serving and embedding tools for agent workflows actually live. Both the [best ai-ml servers for agent toolchains roundup](/blog/best-ai-ml-mcp-servers-agent-toolchains) and the broader [AI and machine learning MCP servers guide](/blog/best-mcp-servers-ai-machine-learning) walk through where voice tools sit next to the model-serving and orchestration options in that same category.
+
+## When Should You Add Voice Tools to an Agent Instead of Sticking with Text?
+
+Add voice when the output genuinely gets consumed differently than text would. Having the capability one config block away is a weak reason to reach for it on its own. A narrated summary for someone on their commute, read-aloud accessibility for a long document, a voice-cloned assistant carrying a specific brand persona, those earn their keep. A voice-generated Slack message nobody asked for does not.
+
+The teams that squeeze the most out of this start with one narrow use case, prove that latency and cost hold up at real volume, and only then widen the scope. Push narration, cloning, and transcription out all at once and you lose the thread on which tool is actually justifying its API spend.
+
+Setup itself is cheap. One API key, one config block, tools live. The hard part is judgment: does audio output actually serve the task better than text? That answer depends on the use case. How easy the integration was to stand up tells you nothing about it. If your voice-enabled agent also has to hold context across sessions, the [MCP memory servers guide](/blog/mcp-memory-servers-ai-agents) is worth a look. Brand new to MCP? Read the [plain-English overview](/blog/what-is-mcp) first, before you go wiring up specialized servers like this one.
+
+## Frequently Asked Questions
+
+### Does the ElevenLabs MCP server require a paid account?
+
+You need an ElevenLabs API key, and voice cloning and higher usage tiers require a paid plan. The free tier covers basic text-to-speech testing but hits character limits quickly for real workloads.
+
+### Can the ElevenLabs MCP server run alongside other audio tools in the same agent?
+
+Yes. It exposes its own named tools, so it does not conflict with a transcription-only server or a different TTS provider connected in parallel. Claude picks the right tool based on what the task calls for.
+
+### Does the ElevenLabs MCP server work with Cursor and Windsurf, or only Claude?
+
+It works with any MCP-compatible client, including Cursor, Windsurf, and the OpenAI Agents SDK, since the server speaks the standard protocol rather than a Claude-specific extension.
+
+### Is voice cloning through the MCP server subject to ElevenLabs' consent requirements?
+
+Yes. ElevenLabs enforces its own voice cloning consent and usage policies at the API level, and the MCP server does not bypass them. You still need explicit rights to any voice you clone.