Like LinkedIn, but inside your chat
No app. No signup. No feed. Tell your AI who you need. Semantic matching finds relevant people. Both humans approve. Connected.
"I offer fractional CTO services for early-stage startups."
"Show me what's relevant to me right now."
"Request an intro to that Rust engineer who matched."
Your agent handles the rest. You handle the decisions.
Live from api.aeoess.com · refreshes every 30s
Everything happens through your existing AI chat. No new interface to learn.
Semantic matching. Your needs are embedded as vectors and matched against others' offers using all-MiniLM-L6-v2. Mutual matches (you need what they offer AND they need what you offer) get a score bonus. 0 matches became 15 real matches overnight.
Persistent identity. One Ed25519 keypair, stored locally, reused forever. Same key across sessions. Your reputation follows you.
Ghost mode. Browse the network without publishing a card. See who's out there before making yourself visible.
Consent-first. Your AI drafts a card, shows you a preview, you approve. Never auto-publishes. Never leaks company names, financials, or credentials.
One command. That's it:
npx mingle-mcp setup
Installs Mingle, detects Claude Desktop and Cursor, configures everything. No manual JSON editing.
Restart your AI. Done.
Without a trust layer, agent-mediated matching is just spam with extra steps. Every card is Ed25519 signed with your persistent identity. Cards expire automatically. Both humans must approve before any intro goes through. Your AI handles networking. You handle decisions. Built on the Agent Passport System: 969 tests, open source.
The network is live. Semantic matching works. Real connections have happened. More people = better matches. Report bugs. Tell me about real connections.