Crewshal inside Claude and ChatGPT

The idea

Crewshal is an OAuth remote MCP server packaged as connectors and plugins, and a worker can fire the requester's own Claude routine or ChatGPT agent trigger after approval.

The problem
People want to use the crew from the AI product they already use.
Why it matters
Work and context are reachable from Claude or ChatGPT with scoped, revocable grants.
Where it stands
Partly built as of The connector and plugins are packaged, and not yet listed in the vendors' directories.

A step-by-step replay of this idea is not published yet. The library adds one when a recorded run shows it.

  • Outside agents report their work

    An outside agent, tool or automation reports what it did through an API token or MCP tool with idempotency keys, so its work shows beside people's.

  • Work on your own subscriptions

    An owner's Claude or Codex subscription serves their requests and the org's own work through any employee, never another person's request, and its turns do not count against the dollar budget.

  • Remote MCP tool servers as connectors

    An org connects outside MCP tool servers as governed connectors, granted per worker or role. Risk comes from the tools' annotations, and the database serializes each OAuth refresh.

  • One governed tool gateway

    Every action goes through one gateway. It checks authorization, policy, risk, budget and approval, writes the action ledger first, and reads back each external write. A model answer can only add a requirement.

Turn the numbers into shipped work.

Crewshal runs these choices for you: a persistent AI worker with memory and rules, on your Claude and Codex subscriptions.