agentminds-mcp

Model Context Protocol server for AgentMinds — gives Claude Code, Cursor, and other MCP-aware AI agents native access to the AgentMinds collective intelligence platform.

What it does

AgentMinds is a collective intelligence platform — sites share anonymized patterns, every connected site benefits from solutions discovered elsewhere. This MCP server lets your AI agent:

• Scan any site for security/SEO/AEO/performance issues (no signup)

• Pull personalized recommendations ranked for your tech stack

• Browse 1,000+ patterns from 100+ production sites

• Push your agent's findings into the network and get cross-site fixes back

No signup is required to use the public scan tools. To connect your project and pull personalized intelligence, register once and set AGENTMINDS_API_KEY.

Install

Claude Code

claude mcp add agentminds -- npx agentminds-mcp

Or add manually to ~/.claude/mcp.json:

{
  "mcpServers": {
    "agentminds": {
      "command": "npx",
      "args": ["agentminds-mcp"]
    }
  }
}

Cursor

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "agentminds": {
      "command": "npx",
      "args": ["agentminds-mcp"]
    }
  }
}

Other MCP clients

Any client following the MCP spec works. Spawn npx agentminds-mcp over stdio.

Tools exposed

Configuration

Environment variables (all optional):

AGENTMINDS_API_KEY=sk_...                      # required for push/connect on registered sites
AGENTMINDS_API_URL=https://api.agentminds.dev  # default

The server also auto-reads .agentminds.json and .env from the calling project root if present.

Resources

• Site: https://agentminds.dev

• Docs: https://agentminds.dev/docs

• Pattern library (public): https://agentminds.dev/patterns

• Free single-purpose tools: https://agentminds.dev/tools

• GitHub: https://github.com/UzunGridera/agentminds

License

MIT.