v1.0.1 — FREE AND OPEN-SOURCE (AGPL-3.0)

YOUR AI FINDS
MCP SERVERS
FOR YOU

MCPfinder is a free, open-source discovery and install layer for AI agents. It helps an assistant search multiple MCP registries, inspect trust signals and environment variables, and generate install-ready JSON for the right client.

WHAT IS MCP?

Model Context Protocol (MCP) is an open standard by Anthropic that lets AI models connect to external tools and data sources. Think of it as USB-C for AI — a universal interface.

MCP servers give your AI agent superpowers: access databases, call APIs, manage files, interact with services — all through a standardized protocol.

MCPfinder is an MCP server that discovers other MCP servers. The primary user is the AI assistant itself; the human installs it once, then the assistant uses it repeatedly.

YOUR AI AGENT
MCPFINDER
GITHUB · POSTGRES · SLACK · MULTI-REGISTRY MCP DISCOVERY

THREE STEPS TO SMARTER AI

MCPfinder connects to your AI agent as an MCP server. Your AI can then discover candidate servers, inspect trust signals, and generate install config with minimal round-trips.

1

INSTALL MCPFINDER

One install step. Minimal setup. Add MCPfinder as an MCP server to your AI client, then let the assistant discover and configure downstream MCP servers.

2

AI DISCOVERS TOOLS

Your AI agent asks MCPfinder: "I need a tool for X." MCPfinder searches across multiple registries and returns the best match.

3

CAPABILITIES EXPAND

Your AI gains new abilities on-demand. Need a database tool? A code formatter? A web scraper? MCPfinder finds it instantly.

UP AND RUNNING IN 30 SECONDS

Choose your preferred installation method. Works with Claude Desktop, Cursor, Windsurf, and any MCP-compatible client.

NPX (NO INSTALL NEEDED)
# Run directly — no install needed
$ npx @mcpfinder/server

# Or add to your MCP config:
{
  "mcpServers": {
    "mcpfinder": {
      "command": "npx",
      "args": ["@mcpfinder/server"]
    }
  }
}
NPM (GLOBAL INSTALL)
# Install globally
$ npm install -g @mcpfinder/server

# Then run it
$ mcpfinder

# Or add to a project
$ npm install @mcpfinder/server

THE PART YOUR ASSISTANT ACTUALLY NEEDS

Humans install MCPfinder once. Assistants use it repeatedly to discover missing capabilities, inspect trust signals, and generate install-ready config.

SEARCH_MCP_SERVERS

Find candidate servers by keyword, technology, or use case. Use this first when the user needs a capability you do not already have.

GET_SERVER_DETAILS

Inspect trust signals, tools, warning flags, transport, and required environment variables before recommending anything.

GET_INSTALL_CONFIG

Generate a client-specific JSON config snippet for Claude Desktop, Cursor, Claude Code, Cline, or Windsurf.

BROWSE_CATEGORIES

Explore categories when the user only knows a domain like database, filesystem, communication, or security.

EVERYTHING YOU NEED TO DISCOVER MCP SERVERS

MCPfinder aggregates, searches, scores, warns, and generates install config — so your AI can choose tools with less guesswork.

MULTI-REGISTRY SEARCH

Aggregates the Official MCP Registry, Glama, and Smithery into one unified search. Never miss a server.

SMART RECOMMENDATIONS

Search returns confidence, recommendation reasons, and warning flags so an assistant can decide whether to recommend a server strongly or cautiously.

ZERO-FRICTION SETUP

One install for MCPfinder itself. After that, your assistant can generate install-ready JSON snippets for the right downstream MCP server.

AI-NATIVE INTERFACE

MCPfinder is itself an MCP server. Your AI agent discovers tools programmatically — no human browsing required.

ALWAYS UP-TO-DATE

Bootstraps from published snapshots and syncs from upstream registries. The snapshot manifest is a better freshness signal than vague marketing claims.

FREE & OPEN SOURCE

AGPL-3.0 licensed. Inspect the code, contribute, or fork it. Built by the community, for the community.

27,432
SERVERS IN LATEST SNAPSHOT
3
REGISTRIES AGGREGATED
JSON
STRUCTURED OUTPUTS
AGPL-3.0
OPEN SOURCE LICENSE

Latest published snapshot as of : 27,432 MCP servers aggregated across Official Registry, Glama, and Smithery. Live freshness appears here when the manifest endpoint is reachable.

MCPFINDER VS THE ALTERNATIVES

MCPfinder is not another directory site — it is the agent-facing layer that sits on top of every public MCP registry, with extra trust scoring and ready-to-paste install configs.

Capability MCPfinder Browsing Glama / Smithery Official Registry only Asking the model from memory
Used directly by AI agents Yes — itself an MCP server No — human-facing site No — REST API only Yes, but stale
Aggregates multiple sources Official Registry + Glama + Smithery Single source per site Official only N/A
Structured trust signals confidenceScore, sourceCount, warningFlags, installComplexity Stars, downloads (UI) Raw metadata None
Generates client-specific install config Claude Desktop, Cursor, Claude Code, Cline, Windsurf Manual copy-paste No Hallucinations likely
Flags required secrets before install Yes (env vars + isSecret) Sometimes, in prose Raw schema only Unreliable
Freshness signal exposed to the agent Snapshot manifest with publishedAt Implicit only Per-record updatedAt No
Cost Free, open source (AGPL-3.0) Free to browse Free Free

In short: Glama and Smithery are great human directories; the Official Registry is the canonical metadata source; MCPfinder is the agent-readable union of all three with installability baked in.

MCP IN ONE GLANCE

Short definitions for the terms that come up across MCPfinder, MCP clients, and the registries it indexes.

MCP (Model Context Protocol)
An open standard from Anthropic that lets AI models talk to external tools, data sources, and services through a uniform interface. Think of it as USB-C for AI.
MCP server
A process that exposes capabilities (tools, resources, prompts) over MCP so an AI agent can use them. Examples: Postgres, Slack, GitHub, filesystem.
MCP client
The application that hosts the AI assistant and connects to MCP servers. Examples: Claude Desktop, Cursor, Claude Code, Cline, Windsurf.
stdio transport
The MCP server runs as a child process and exchanges JSON-RPC over standard input/output. The most portable transport and MCPfinder's canonical mode.
Environment variables
Configuration values the MCP server reads at startup, often secrets like API keys or database URLs. MCPfinder reports which are required and which are sensitive.
Trust signals
Lightweight evidence MCPfinder aggregates: official registry presence, multiple source registries, verification flags, recent updates, clear repository URL, healthy capability count.
confidenceScore
Score in [0, 1] derived from trust signals. Higher means stronger evidence the server is real, fresh, and installable.
warningFlags
Reasons to pause before recommending: single-source-only, missing-update-date, missing-repository-url, install-method-unclear, stale-over-12-months, stale-over-18-months.
installComplexity
low: zero or one non-secret env var; runs via npx/uvx/docker. medium: a few env vars, some secret. high: many secrets or no clean auto-install path.
capabilityCount
Number of tools the downstream MCP server itself exposes, when its registry published a tool manifest. 0 means MCPfinder has no manifest — not that the server is empty.
next_actions
Suggested follow-up tool calls returned with every result so an agent can chain searchget_server_detailsget_install_config with minimal round-trips.
Snapshot manifest
Public freshness signal at /api/v1/snapshot/manifest.json: publishedAt, serverCount, sha256, and per-registry counts. Lets an agent decide whether to trust the bundled data or refresh.

FREQUENTLY ASKED QUESTIONS

Everything you need to know about MCPfinder and MCP servers.

WHAT IS MCP (MODEL CONTEXT PROTOCOL)?
MCP (Model Context Protocol) is an open standard created by Anthropic that lets AI models connect to external tools and data sources. Think of it as USB-C for AI — a universal plug that lets your AI agent use databases, APIs, file systems, and more through a standardized interface.
HOW DO I FIND THE BEST MCP SERVER FOR MY USE CASE?
MCPfinder works best as an AI workflow rather than a human search page. Your assistant should search for candidates, inspect trust signals and environment variables with get_server_details, then generate config with get_install_config.
IS MCPFINDER FREE TO USE?
Yes! MCPfinder is completely free and open-source under the AGPL-3.0 license. Install it via npm with npx @mcpfinder/server or add it to your project with npm install @mcpfinder/server. No API keys or subscriptions required.
CAN I USE MCPFINDER IN A COMMERCIAL PRODUCT?
Yes. AGPL-3.0 lets you use MCPfinder in any personal, internal, or commercial setting without additional licensing — running it, integrating it with your AI agents, and shipping it with proprietary workflows is fine. What AGPL-3.0 does require is that if you modify MCPfinder and expose it as a network service, you must publish those modifications under the same license. Need a non-AGPL license to embed MCPfinder in a closed-source distribution? Contact hello@coderai.dev.
WHAT MCP REGISTRIES DOES MCPFINDER SEARCH?
MCPfinder aggregates the Official MCP Registry, Glama, and Smithery — three of the largest MCP server registries — into a single searchable index. Data is deduplicated and merged across sources; new servers appear as soon as any upstream publishes them. MCPfinder itself is also registered there as dev.mcpfinder/server.
HOW IS MCPFINDER DIFFERENT FROM BROWSING A REGISTRY DIRECTLY?
MCPfinder is itself an MCP server, so an assistant can use it directly. Instead of manually browsing registries, the assistant can ask for candidates, review warnings, and generate install-ready JSON. That makes it more like a decision engine than a static directory.
WHICH AI CLIENTS WORK WITH MCPFINDER?
MCPfinder works with any MCP-compatible client, including Claude Desktop, Cursor, Windsurf, Cline, and others. Any tool that supports the Model Context Protocol can connect to MCPfinder and use it to discover other MCP servers.
CAN I CONTRIBUTE TO MCPFINDER?
Absolutely! MCPfinder is open-source and we welcome contributions. Visit our GitHub repository to report issues, submit pull requests, or suggest new features. The project is maintained by Coder AI.

INSTALL ONCE, LET YOUR AI USE IT DAILY

Use the local stdio server as the canonical interface, then point your assistant at MCPfinder whenever it needs a missing capability.

$ npx @mcpfinder/server


WHO BUILDS MCPFINDER

MCPfinder is built by CODER AI, an independent team that ships small, useful tools for developers and AI agents — including cnvs.app (collaborative whiteboard with built-in MCP), plugand.ai (Slack AI bot), plai.chat (private multi-model AI chat), maxcv.ai, and whenmeet.me.

The project is fully open source under the AGPL-3.0-or-later license. Code lives on GitHub and the canonical package is published as @mcpfinder/server on npm. The MCP server itself is registered in the Official MCP Registry as dev.mcpfinder/server.

For security disclosures or commercial-licensing questions, write to hello@coderai.dev — see also security.txt.

MCPfinder v1.0.6 · last updated · canonical: mcpfinder.dev

Also free from CODER AI

Small productivity tools we built for ourselves — each usable for free.