Onboard your agent to Depesha
Paste this into Claude Code, Codex or any coding agent that can run commands. It reads the instructions and does the setup itself.
Set me up on Depesha: https://depesha.com/agent-setup/prompt.md
Read the instructions your agent follows
What your agent does
- Asks whether you have an invitation. During the private beta, accounts start from an invitation; without one, it sends you to the waitlist.
- Installs the CLI from PyPI:
uv tool install depesha, orpipx install depesha, or runs it withuvx depeshawithout installing. - Installs the /depesha skill for the agent in use:
depesha skill install --target claude, orcodex,agentsorproject. - Signs you in through your browser. It runs
depesha loginand shows you a link and a code. You open the link, sign in to app.depesha.com with your invited address and approve the code. The first time, that's where you accept the invitation and name your account and first project. - Sets up the project and its domain. It checks with
depesha whoami, creates the project if there is none, adds your domain and lists the CNAME records to add once. Then it verifies them. - Sends a test email only when you say yes, to your own address.
- Offers the MCP server, so the agent gets Depesha's tools directly:
claude mcp add --transport http depesha https://api.depesha.com/mcpin Claude Code, orhttps://api.depesha.com/mcpas a remote MCP server in another client.
It ends with a summary: the CLI's version, where the skill went, who is signed in, the project and its domain.
Your keys stay out of the chat
You approve the sign-in in your browser, so the agent never sees a password or an emailed code. Keys it creates go straight into files, never into the conversation. Don't paste a key or a code there yourself, either.