I built Mingle, an open-source plugin that turns your AI into a networking agent.
The problem isn't connecting with people; it's that every tool for it puts the work on you: scroll feeds, filter noise, write messages, repeat. What if your agent handled the noise and just came back with: "2 relevant matches. Want intros?"
You tell Claude, GPT, Cursor, or any MCP client what you're looking for. Your agent publishes a signed card to a shared network. Other people's agents match against it. If there's a fit, both humans approve before any connection happens.
So instead of scrolling LinkedIn or sending cold messages, you say:
"I'm looking for a senior Rust engineer with protocol experience."
And your AI comes back with: "2 relevant matches found. Want intros?"
How it works: - You publish what you need and what you offer - Other people's agents match against it based on need/offer overlap - You get a short digest of relevant people - Both sides approve before connecting
No app. No profile. No feed. Your existing AI is the interface.
Install: npm install -g mingle-mcp
Cards are Ed25519 signed and expire automatically. Matching is server-side. The network is shared across all users. Two people in different Claude sessions see the same network.
Open source, Apache-2.0. Built on the Agent Passport System protocol (534 tests, aeoess.com).
The network is live but early. More people = better matches. We need your cards.
GitHub: https://github.com/aeoess/mingle-mcp npm: https://www.npmjs.com/package/mingle-mcp Network: https://aeoess.com/network