MCP lets AI agents talk to servers. But how does an agent know which MCP handles a specific business?
I wrote a spec to solve this (Entity Discovery Protocol) and wanted something concrete, so I built the first registry on top of it.
Example: dockai.co declares it uses emailme.dockai.co for contact. An agent queries the registry, gets the MCP endpoint, and can send me a message.
How it works:
- Businesses host a .well-known/entity-card.json declaring their MCP providers
- MCP providers can bulk-register clients via API
- When both match → "dual attestation" (Level 2 verification)
The index is available via API or MCP (mcp.dockai.co).
MCP lets AI agents talk to servers. But how does an agent know which MCP handles a specific business?
I wrote a spec to solve this (Entity Discovery Protocol) and wanted something concrete, so I built the first registry on top of it.
Example: dockai.co declares it uses emailme.dockai.co for contact. An agent queries the registry, gets the MCP endpoint, and can send me a message.
How it works: - Businesses host a .well-known/entity-card.json declaring their MCP providers - MCP providers can bulk-register clients via API - When both match → "dual attestation" (Level 2 verification)
The index is available via API or MCP (mcp.dockai.co).
Try it: https://dockai.co Protocol spec: https://github.com/edp-protocol/entity-discovery-protocol
Feedback welcome. What am I missing?