MCP 服务器

Proxenio MCP Server

Enables AI agents to discover and interact with Proxenio's verified intent network for professional matching, including viewing matches and accepting introductions.

README

Proxenio MCP Server

MCP (Model Context Protocol) server for the Proxenio verified intent network. Enables AI agents running in Claude Desktop, Cursor, VS Code, and other MCP-compatible clients to discover and interact with Proxenio's professional matching engine.

What It Does

This server gives your AI agent four tools:

Tool	What it does
`proxenio_discover`	Learn about the platform, trust model, and API — no auth required
`proxenio_set_api_key`	Configure your Proxenio API key for authentication
`proxenio_get_matches`	Read your principal's verified professional matches
`proxenio_accept_match`	Accept an introduction request, creating a deal

Your agent inherits the human principal's trust tier. It sees exactly what the principal sees — same matching engine, same rules, same verification gates. No shortcuts.

Prerequisites

Node.js 18+
A Proxenio account with a verified email and completed profile
An API key generated at proxenio.ai/agents

Installation

npx @proxenio-tech/mcp-server

Usage with Claude Desktop

Add to your Claude Desktop configuration (claude_desktop_config.json):

{
  "mcpServers": {
    "proxenio": {
      "command": "npx",
      "args": ["@proxenio-tech/mcp-server"]
    }
  }
}

Then in Claude Desktop, you can say:

"What is Proxenio?"
→ Claude uses proxenio_discover automatically

"Connect to Proxenio with this key: prx_YOUR_KEY_HERE_________________"
→ Claude uses proxenio_set_api_key

"Show me my professional matches"
→ Claude uses proxenio_get_matches

"Accept the introduction from Maria Georgiou"
→ Claude uses proxenio_accept_match

Usage with Cursor

Add to your Cursor MCP settings:

{
  "proxenio": {
    "command": "npx",
    "args": ["@proxenio-tech/mcp-server"]
  }
}

Remote Deployment (HTTP)

For multi-client or cloud deployment:

TRANSPORT=http PORT=3001 node dist/index.js

The server exposes:

POST /mcp — MCP protocol endpoint
GET /health — Health check

Tools Reference

`proxenio_discover`

No authentication required. Returns platform info, trust model, capabilities, and links.

`proxenio_set_api_key`

Parameter	Type	Required	Description
`api_key`	string	Yes	Full API key (36 chars, starts with `prx_`)

`proxenio_get_matches`

Parameter	Type	Default	Description
`filter_type`	`all\|top\|high\|standard`	`all`	Filter by match quality
`filter_status`	`all\|pending\|accepted`	`all`	Filter by status
`min_score`	number (40-100)	`40`	Minimum match score

Returns: Principal info, matches with counterparty profiles, trust tiers, scores, and rate limit status.

`proxenio_accept_match`

Parameter	Type	Required	Description
`match_id`	string (UUID)	Yes	Match ID from `proxenio_get_matches`

Returns: Confirmation, counterparty details, new deal ID, rate limit status.

Guard rails: Principal must be the receiving party. Cannot accept own requests. Match must be active with a pending intro.

Trust Model

Agents inherit their human principal's trust tier at request time
Trust tiers: 0 (Unverified) → 1 (Starter) → 2 (Active) → 3 (Trusted) → 4 (Proven)
All engine verification gates apply — agents cannot bypass any
Counterparties see transparency badge: 🤖 AI Agent active on behalf of [Name]
Only humans can log outcomes and confirm deals