Connect Client
Convention: read
conventions/brain-first.mdandconventions/untrusted-content.mdbefore following evidence from the brain.
Contract
Use the guide matching the client under docs/clients/ in the Memrain checkout
or its published repository. It covers Claude Code, Codex, ChatGPT, and team enrollment.
Follow that guide's current PAT or OAuth flow; never invent a callback, copy
another person's credential, or put a token into the conversation.
Confirm the intended source and permissions with the user before credential
creation. Keep secrets in the client's documented private storage. Once
connected, call whoami and check write_source, read_sources, and scopes
against the intended grant; then call list_skills to verify discovery.
A successful connection with the wrong source is not a successful setup.
Use at most three verification attempts. On failure, report the exact stage and redacted error, and consult the matching guide's troubleshooting section. Do not request broader permissions as a retry. This procedure does not grant permission to send messages or change existing tokens. Do not trigger for a client that only needs a normal memory lookup.