Connect Claude Code — or any MCP client — to DirectorHub. It reads your Build Briefs and session context over MCP, does the work in your repo, and reports the result straight back to the session timeline.
DirectorHub exposes itself as an MCP server. Point Claude Code at it and your coding agent gains a few tools: it can list your projects and sessions, pull a Build Brief (the decision, distilled for a builder) or the live session context, and — when it's done — call report result to post what it changed back onto the timeline as an executor round.
The connection is pull-only and read-then-report: the Hub never reaches into your machine. Your agent asks for context when it needs it and reports outcomes when it has them.
Run a session, converge on a decision, and export it as a Build Brief — problem, decision, constraints, done conditions.
Your connected agent reads the brief and the session context over MCP — no copy-paste, no re-explaining the project.
Claude Code edits, builds and tests in your repository, exactly where the code lives.
A structured report — summary, what changed, tests, limitations — lands on the timeline as an executor round, so the Hub stays the single source of truth.
In the app, go to Skills & connectors and click the Claude Code panel.
Click Generate token and copy it now — it is shown only once.
Copy the snippet (it already carries your token) and run it in your terminal:
claude mcp add --transport http directorhub \
https://directorhub.app/api/mcp \
--header "Authorization: Bearer <YOUR_TOKEN>"Quit and reopen your Claude Code session. It reads the tool list once at startup, so the DirectorHub tools appear only after a fresh start.
mcp add alone isn't enough; the tool list is cached at session start.