Introduction
Describe the app in plain words: purchase approvals, a parts inventory, spa bookings. The agent you already use (Claude Code, Claude, ChatGPT, Gemini, Cursor, Codex) writes the code on the Manh framework over MCP. manh.run checks it, previews it and puts it online at <id>.manh.run with its own UI, API and SQLite database.
Agent prompt
Paste this into Claude Code, Claude or ChatGPT. The agent connects to manh.run, looks at what you have and suggests what to do next.
Help me get started with manh.run. Do the following:
1. Check whether the "manh" MCP server is set up. If you are Claude Code and it is not, run
claude mcp add --transport http manh https://mcp.manh.run
then tell me to type /mcp → pick manh → Authenticate.
Other agents: walk me through adding https://mcp.manh.run
(steps: https://manh.run/en/docs/getting-started/connect).
2. Call list_projects and tell me which projects I already have.
3. Call framework_docs("process") to learn how apps are built.
4. Ask me what app I want (for example: purchase approvals, inventory, spa bookings).
Write spec/ first, wait for my review before coding, and run check after each phase.
5. When done, suggest 3 next steps that fit the app: sign-in, email,
a scheduled job, or deploying it live.Using Claude, ChatGPT or Gemini on the web? See Connect your agent, with a screenshot for every step.
I want to…
- Build create, read, update, delete A database table, API and a list page with a form for one object, such as orders.
- Let users sign in Turn on sign-in with
app/auth.tsand set per-role access for each API. - Send email from the app Send email from templates the owner can edit, and take replies back into the app.
- Run work on a schedule Cancel stale orders every hour or send a report at 8 AM with one cron file.
- Add your own domain Serve the app on your domain instead of
<id>.manh.run. - Deploy and roll back Put the app online once
checkpasses, roll back when a release breaks. - Build a dashboard Each role’s to-do list and a few key numbers on one screen.
- Call another system’s API Pull data from software you already run, with the key kept on the server.
Docs by section
- Get started Read in order: connect your agent, first app, pages, styling, API, database, sign-in, check, deploy.
- Structure What each file in
app/does:page.tsx,get.ts,middleware.ts,*.table.ts… - Guide Concepts, how to build each kind of screen, the CRUD recipe, working with agents.
- API Hooks, server functions, MCP tools,
checkrules and plan limits. - Deploy Put the app on
<id>.manh.run, add your own domain, read logs, roll back. - Examples Business screens that run right in the page, with code and prompt.
Next steps
- Connect your agent — add
https://mcp.manh.runto the agent you use, about 2 minutes. - Your first app — from a business description to a running preview, with the spec your agent writes.
- Briefing your agent — write the request and review the spec so the agent gets it right the first time.
Last updated