AI agents & the MCP server¶

Specivo ships with a built-in MCP server — Model Context Protocol, the open standard that AI clients use to talk to outside tools. With it, an AI assistant can work inside your tracker: read an issue, search the wiki, leave a comment, log time, or update status — instead of you copying text back and forth.

Any client that speaks MCP can connect: Claude Code, Codex, Cursor, Windsurf, Cline, and others. The agent authenticates with an API key tied to a real Specivo user, so it only sees and changes what that user is allowed to, and every write it makes is recorded.

How it works¶

An AI client connects to your Specivo server's MCP endpoint using an API key. Once connected, the client discovers a catalog of around 40 tools (all prefixed specivo_) and calls them on your behalf when you ask it to. You stay in control: you decide which client to connect, which account's key it uses, and you can cut off access at any time.

Two things make this safe to hand to an agent:

• Permissions are enforced. The agent acts as the user behind the key. A key from an account with no write role on a project can read but not change it. See API keys & scopes.
• Everything is audited. Every write goes to the security audit log tagged source=mcp, so a project manager can see exactly what an agent did and when. See What agents can do.

Two endpoints¶

The MCP server exposes the same tools over two transports. Pick whichever your client supports:

Both serve the identical tool catalog — the difference is only how the client talks to the server. See Connecting an AI client for the exact configuration each one needs.

Where to go next¶

• API keys & scopes — create a key with the mcp scope, and why teams use a dedicated service account.
• Connecting an AI client — copy-paste config for SSE clients and Codex.
• What agents can do — the tool catalog by area, read vs write, and auditing.
• Common workflows — concrete things to ask your agent to do.