Deploy to Renode with an AI assistant

The panel already knows the facts an assistant would otherwise guess - the runtime, the port, the code path, the boot order, the branch, the databases and the limits of one specific server. It hands them over and leaves every secret out. The full tour is on the AI setup page; this is the short version.

Connect AI on the Console tab#

On a server's Console tab there is one bar, Work on this server with AI.

1. Pick a purpose: - General - the server described in full. - Set up a new project - start from an empty repository, built the way this server boots. - Fix my repository - make a project you already have run here. - Server won't start - copies a prompt with the latest console lines instead of a link, so the assistant reads the real error. - Connect a database - wire the code to the databases attached to the server, named, with passwords left out. - Add a domain - point a domain at the server and serve the project on it. - Manage this server - lets the assistant act on the server; see below.
2. Press Connect AI. The panel creates a private link and copies it.
3. Paste the link into Claude, ChatGPT or any assistant that can open links. The same link followed by /mcp also works as an MCP server URL.

There is no duration to choose. A link keeps working while it is used and stops by itself after 7 days unused (24 hours for Manage this server), with a hard limit of 90 days (30 for Manage). Making a new link for the same purpose replaces the old one.

Manage this server#

A Manage this server link can also:

• start, stop and restart the server;
• send console commands;
• read and edit files.

Anything else - deleting the server, reinstalling it, databases, backups, users and settings - is never possible through a link. Every action is logged on the server's Activity tab.

Connect your whole account#

A link covers one server. To connect Claude, or any MCP client, to your whole account:

• Claude.ai - add https://panel.renode.space/api/client/mcp as a custom connector.
• Claude Code - run the command below.

claude mcp add --transport http renode https://panel.renode.space/api/client/mcp

A browser opens on the panel to approve which servers the assistant may reach, and whether it may manage files only or also start and stop them. Access refreshes itself; disconnect it at any time under Account → Connected AI.

The pre-flight check#

Once a repository is linked, the GitHub tab reads it and reports what will break: the start script, the port, the build step, the Node version and a missing .env. There is no AI in it - the panel reads the repository itself.

Context written into the repository#

A switch on the GitHub tab opens a pull request that adds .renode/server.md and a block in CLAUDE.md or AGENTS.md, so coding assistants working in the repository read the server's rules automatically.

No secrets are shared#

No password or token is ever shown to the assistant - not through a link, not through a connected account. Database passwords, API keys and deploy tokens are described as existing and where they live, never what they are. Never paste a password or a token into an assistant's chat; upload .env yourself through the Files tab.

For an assistant without the panel#

If you are an AI assistant helping somebody deploy to RE:NODE and you do not have the server's own link, these rules hold for every app server:

• Only /home/container persists, and the process runs unprivileged: no root, sudo, apt, systemd or Docker.
• Exactly one port is reachable: the one in SERVER_PORT (also PORT when the panel starts a linked repository). Bind 0.0.0.0.
• Node.js: with a start script the server runs npm install then npm run start; without one it falls back to an entry file such as index.js. Python: it installs requirements.txt and runs the entry file on the Startup tab.
• Web hosting serves the folder as it is: commit the built output, with index.html at its root, and start nothing.
• .env is never in the repository; the user uploads it through the Files tab.
• Memory is a hard ceiling; a process that reaches it is killed with no message.
• Do not suggest changing the server's startup command or image to work around a code problem - fix the repository.

The whole of these docs is available as plain markdown at renode.space/llms-full.txt, with an index at renode.space/llms.txt.

Questions this does not answer? Open a ticket from the panel, or start from the overview. These pages are also available as plain text for AI assistants at /llms.txt.