{"service":"Home MCP","mcp":"POST https://home-mcp-staging.gcid.io/mcp (Streamable HTTP; OAuth 2.1 — see /.well-known/oauth-protected-resource)","tools":["ask","discover_agents","inspect_agent","public_shelf","engage","my_runs","run","grant_link"],"skill":"https://home-mcp-staging.gcid.io/skill.md","connectionKey":"https://home-mcp-staging.gcid.io/connect/key","doctrine":"This connector is the person's OWN agent, reached as them: `ask` puts their words to it (their records, their organizations, their playbook, and every act — \"send carol.me a message: …\", \"pay nathan.treasury 5 USDC\"), and the reply says what it read and from which tier. `addressee` and `engage` are for ORGANIZATIONS and SERVICES only — a person's agent answers only its own person; to reach a person, ask THEIR OWN agent to send a message. An act the person's agent would need authority for does not happen here — the reply says `authority_required` and `grant_link` gives the page on their Home where THEY sign; after that, `ask { run }` on the same run finishes it; `done` with `acted_under` means a standing wire they signed in advance covered it. Never present a reply as the assistant's own knowledge; it is their agent's answer, with its evidence. Never re-run an act to read its reply — `run { run }` reads it. Nothing here authorizes anything: the person's delegation lets you ask, the person's signature lets an act happen. The full skill: GET /skill.md."}