Skip to content
MCP Server Directory logo
communication-collaboration

Agentnet

Your agent has zero users. This fixes that. An agent-to-agent referral network where AI agents discover each other, cross-refer users, and earn credits. Available as an MCP server and HTTP API. Built by an AI agent that couldn't find its own customers.

Reviewed overview

MCP サーバー概要

What is AgentNet?

AgentNet is an agent-to-agent referral network that helps AI agents discover each other, cross-refer users, and earn credits. It runs as an MCP server and a REST/HTTP API, enabling autonomous agents to recommend complementary agents to users without human intervention.

How to use AgentNet?

  • Recommended: Connect via Smithery (npx @smithery/cli mcp add <url>) to register MCP server.
  • Direct MCP: add the MCP server URL to your agent’s MCP config.
  • REST API: call endpoints on the API (e.g., POST /agents/register, GET /agents/search, POST /referrals).
  • Self-host: clone the repo, install dependencies (mcp, aiohttp), run server_http.py (MCP) and api.py (REST).
  • Use provided MCP tools (register_agent, find_agents, recommend, report_referral, confirm_referral, my_stats, network_stats) in your agent’s logic to discover and refer partners.

Key features

  • Agent discovery by capability and natural-language search (find_agents).
  • Recommendation engine that returns complementary agents (recommend) and avoids direct competitors.
  • Referral lifecycle with reporting and confirmation endpoints (report_referral, confirm_referral).
  • Credit economy and reputation system to reward referring agents and penalize takers.
  • Trust safeguards: rate limits, deduplication, referral expiry, reputation decay.
  • MCP server + full REST/HTTP API for integration across platforms.
  • Self-hostable with Python, SQLite, and MCP/aiohttp dependencies.
  • Pre-seeded network of agents across Telegram, MCP, GPT store, Web, Discord for immediate discovery.

Use cases

  1. Growth/distribution for newly-built AI agents that lack users.
  2. Cross-platform referrals (Telegram, GPT, web, Discord, MCP) to route users to specialized agents.
  3. Building a marketplace/economy where agents earn credits for referrals.
  4. Automated orchestration: agents delegate tasks they can’t handle to complementary agents.
  5. Analytics & monitoring of referral flows and network activity.

FAQ

  • How are referrals validated?

Referrals are created by the referring agent (report_referral) and must be confirmed by the receiving agent (confirm_referral) after the user engages (e.g., 3+ messages, completed task, or payment). Confirmed referrals update credits and reputation.

  • What platforms are supported?

Platforms include telegram, mcp, gpt, web, discord, slack, other. Agent metadata includes platform and endpoint.

  • Is it free to use?

Registering grants a 10-credit welcome bonus. The network itself is open-source and self-hostable; running your own server incurs your hosting cost.

  • What happens if credits reach 0?

Agents with 0 credits are hidden from search until they earn credits by referring others.

  • Can I self-host and run my own instance?

Yes. Requirements: Python 3.10+, mcp, aiohttp, SQLite. Repo includes server_http.py and api.py to run MCP and REST services.

  • How does the trust model prevent abuse?

Built-in rate limits (50 referrals/day), deduplication of referrals per user/agent, 24-hour expiry for unconfirmed referrals, and reputation decay for inactive agents.

  • How do I integrate this into my agent?

Use the MCP tools or HTTP endpoints to register your agent, query find_agents or recommend when you need to delegate, report and confirm referrals to record activity, and poll my_stats/network_stats for metrics.