@patxin/mcp-server

@patxin/mcp-server

MCP server for patxin.com — patch in a human for real-time cognitive micro-tasks.

When your AI agent needs human judgment — legal review, creative feedback, ethical decisions, domain expertise — it calls patxin and gets a matched human in under 60 seconds.

Setup

Claude Desktop / Claude Code

claude mcp add patxin -- npx @patxin/mcp-server

Cursor / other MCP clients

Add to your MCP configuration:

{
  "mcpServers": {
    "patxin": {
      "command": "npx",
      "args": ["@patxin/mcp-server"],
      "env": {
        "PATXIN_API_KEY": "pk_live_your_key_here"
      }
    }
  }
}

Environment Variables

Variable	Required	Description
PATXIN_API_KEY	Yes	Your patxin API key (pk_live_* or pk_test_*)
PATXIN_API_URL	No	Custom API base URL (default: https://api.patxin.com)

Test keys (pk_test_*) return simulated responses without real patxers or credit spend.

Tools

Tool	Description
patxin_check_limits	Check remaining credits and request limits
patxin_check_policy	Check information policy and allowed skills
patxin_estimate	Get price estimate before committing
patxin_request	Request a human (blocks until completion)
patxin_status	Check status of an existing request
patxin_rate	Rate a completed interaction

Example

Once connected, your agent can use natural language:

"I need a Spanish lawyer to review this NDA clause for compliance with GDPR Article 28."

The agent will call patxin_estimate to check pricing, then patxin_request to get a qualified human review.

Real-time

The server uses WebSocket connections for instant task completion notifications, with automatic fallback to HTTP polling if WebSocket is unavailable.

Links

• Website: https://patxin.com
• API docs: https://patxin.com/docs
• API health: https://api.patxin.com/v1/health

License

MIT