Connect your AI, keys and services

Build from Claude, Cursor or any AI #

Your space is an MCP server. In HQ → Connections, make an access token, then add it to your AI app. In Claude Code, for example:

claude mcp add --transport http zvessa https://your-backend/mcp --header "Authorization: Bearer mbk_…"

New tokens can read and propose. The outside AI proposes changes, and you approve them in the app. Only give a token "approve" for an AI you'd let change the space on its own (it can still be undone).

Your own keys #

Save keys in Connections. They're sealed on the server, and nobody can read one back, including the AI. Each key only goes to the hosts you allow. A Claude key with use "AI" is what your space's AI runs on. The quickest way to add it is Connect your AI at the top of HQ.

Outside services #

Connect another MCP server or any HTTPS API. The AI and mini-apps can then read from it. They can only read, unless you allow more. Details for developers: Capabilities.