Documentation
Using RepoGo
MCP servers
MCP servers give your agents tools beyond your code, like your issue tracker, database or docs.
Add a server#
- In a chat, tap + and choose MCP Servers.
- Tap Add MCP server, then pick a preset or enter the server's URL.
Presets include Context7, DeepWiki, Cloudflare Docs, Linear, Stripe, Vercel, Notion, Sentry, Supabase, Neon, PostHog, and Jira & Confluence.
Either one opens the Connect sheet. A server with no setup connects at once. Otherwise it asks for an API key, or opens the provider's sign-in page and comes back to RepoGo when you approve.
A server that's already connected#
Picking a server you've already connected shows it as Connected, with:
- Sign In Again, for a sign-in that expired or was revoked. The server keeps its name and projects.
- Add Another Account, to connect a second account on the same server, such as a work and a personal Linear. Each account is its own server in the list, which you can rename; turn on the one you want for each project.
- Turn Off for This Project, which leaves it connected for your other projects.
When an agent needs a server#
An agent can ask you to connect a server, an API key or a CLI sign-in with a Connect card in chat. The card's button says whether it's connected, and tapping it opens the same Connect sheet. Keys and sign-ins never pass through the chat: what you type goes straight to the environment, and the agent only learns that it's connected.
Choose servers per project#
Each project has its own toggles, so a server is only available to the agents in the projects where you turn it on.
Use a server in cards#
Cards can show live data from the servers you turned on for a project. Ask your agent for one, like "a card listing my Neon branches". The card loads when it appears, and pulling down on it loads it again.
Buttons on a card can run a server's tools too. A tool that could delete or overwrite data asks you before it runs, and the card's data reloads afterwards.
Only servers you added in RepoGo work in cards. Servers set up in Claude's or Codex's own settings don't.
The built-in RepoGo server#
RepoGo adds its own server, which lets your agent drive the in-app browser and run your project's plugins. It also teaches your agent to answer with cards, the native screens and buttons RepoGo draws in chat.
The server can also reach your chats on the environment, in every project. It works on them the same way the chat drawer does: list and search them, read them, rename, pin, resolve or stop them, and send them messages. Ask things like "rename my last five chats by what they did" or "tell the chats about the login bug to rerun their tests". An agent can't send a message to its own chat.
Cards can show your chats too, like "a card of my open chats". The card updates as chats change. Tapping a row opens that chat, and its buttons can pin, resolve or message one.
It's on for every project on an environment. To run plain Claude or Codex there, with no RepoGo tools and nothing added to its instructions, switch RepoGo off under Built In; that turns it off for the whole environment.
Registering your own sign-in app#
Some providers make you register an OAuth app yourself. Set its redirect URI to:
https://repogo.app/mcp/callback/native