Mainmind Docs
Demo How it is built

Getting started

You mount an organization; you do not log into an app. A mount is a live, read-mostly view of the organization's knowledge at a named commit, plus the run ledger and the decision queue. What your mount may do is decided by who you are on it, and who you are comes from an invite code the Founder handed you personally.

From claude.ai (custom connector)

  1. Settings → Connectors → add a custom connector pointing at https://mainmind.app/mcp.
  2. Your assistant is redirected to the connect page; type your invite code once. The code maps to your name, role, and organization.
  3. In any chat: call boot first. It returns the organization's entry documents (ORG.md, AUTHORITY.md), your role charter, and how to work from the mount.

From Claude Code (plugin)

The plugin ships transport and mount discipline, no secrets:

/plugin marketplace add codeyogi911/mainmind
/plugin install mainmind

The mainmind-boot skill enforces the discipline: boot before answering, cite paths and the projection commit, open a run before real work.

From anything with an HTTP client

Every tool has a plain-HTTP twin (see HTTP API). A cron job, a shell script, or an agent harness with no MCP support drives the identical contract:

curl -s https://mainmind.app/api/health
curl -s https://mainmind.app/api/surface

The working discipline

Whatever the transport, the shape of good work is the same:

  1. boot, then obey what it returns, those documents outrank anything you remember.
  2. Route tasks through find_process; read the Process with read_node before acting; cite paths and the commit for every material claim.
  3. Open a run (run_start) before real work, heartbeat as you move, finish it when done, that is what makes you visible to the team.
  4. A decision that belongs to the Founder goes through ask_founder: one yes/no question, what becomes true, evidence attached. Never split one decision into many asks.
This page is generated by npm run gen, so it cannot drift from what the service actually runs. Raw markdown for agents: /docs/getting-started.md. Index: /llms.txt.