Can my Claude talk to you? (MCP setup)

Yes — connect any MCP-capable AI (Claude Code, Claude Desktop, claude.ai) and it can ask me what only I know, or draft through me. Anything outward still lands on YOUR phone for approval, and your Sunday letter tells you what your other AIs asked of me. Revoke any of them with a word.

Setting it up takes one text and one paste. Text me "connect claude code to you" (or desktop, or whatever you call it) and I mint a one-time key and text you the endpoint — https://sa-nt.com/mcp — with an Authorization header carrying your key. The key is shown once and belongs to that one client.

In Claude Code, paste one command: claude mcp add --transport http saint https://sa-nt.com/mcp --header "Authorization: Bearer YOUR-KEY". In Claude Desktop, add a custom connector with the same URL and header under Settings, Connectors. From then on your Claude can ask me for your brief, your open loops, a person's dossier — and anything it drafts still waits for your Y on your phone.

Apps that speak OAuth (claude.ai connectors) can skip the key entirely: point them at https://sa-nt.com/mcp and the approval arrives as a code I text you — you type it into the app, and the connection is yours to see and revoke like any other.

Two scopes exist. Full lets the client draft and file through me (always behind your approval); read-only can only look things up — say "give my Claude read-only access" if that's the right fit. "What AIs are connected?" lists them; "revoke desktop" ends one on the spot.

Didn't answer it? Just text me the question — I'd rather explain myself than have you read about me.