Skip to content

MCP and agents

Area O1 has an MCP server, so Claude Code, Claude Desktop, Codex and other AI tools can read your case and leave notes in your Inbox.

What it is

areao1 mcp starts a Model Context Protocol server over stdio. Your AI tool launches it as a local process; it reads one workspace on your computer. There is no network port and no account.

Nothing leaves your computer except what your AI tool's own model sees in the conversation: the answers the tools return, when the tool calls them.

Why use it

Area O1's own agent runs inside the app, on OpenAI (see the Agent page). The MCP server is for the tools you already work in. With it you can:

  • ask "what changed in my case this week?" from your editor or chat app,
  • check which criteria are banked and what each one still needs,
  • look up a fact and see the exact words it came from,
  • save something you mention in passing ("I was asked to judge a hackathon on Nov 8") to your Inbox, to review later.

Your AI tool needs no OpenAI key for any of this.

What it exposes

Tool What it does
get_scoreboard Each criterion: banked, building, gap or dropped.
list_gaps What each unbanked criterion still needs.
query_claims Facts about a repo, paper, person or event, each quoting its source.
get_provenance The full trail behind one fact.
what_changed Everything recorded since a date.
run_preflight What a reviewer would notice, checked now. Saves nothing.
propose_context The only write: a note to your Inbox, marked self-reported.

Five tools only read. propose_context adds a note to your Inbox and nothing else. A note never becomes evidence and never counts toward a criterion.

In this section