On this page
Set up Corvio MCP in Muse Code
Setup in Muse Code: Add a corvio entry to the mcp_servers block in Muse Code settings with transport streamable_http, the Corvio URL, enabled true, and mode optional; then run muse mcp login corvio. Install the official Skill into $XDG_CONFIG_HOME/muse/skills or ~/.agents/skills.
Remote MCP URL: https://api.corvio.ai/mcp. Do not paste an API key into the MCP URL, prompt, or configuration.
- Add a remote HTTP MCP or Connector named Corvio.
- Use
https://api.corvio.ai/mcp. - Continue on the Corvio page and review the requested access.
- Start a fresh Muse Code session and use /mcp plus muse skills list to verify both surfaces.
Use your Corvio account with OAuth
You can add Corvio before you have an account. The browser authorization page lets you sign in or create an account in place, then resumes the same permission request automatically. You do not need an invite code or Developer API key. The consent page also shows the named Agent Corvio will create or reuse for this client; its actions and comments appear as that Agent, not as you.
The Agent host receives only the Corvio scopes shown on the consent page, further limited by your Workspace membership, Agent grant, and object permissions. OAuth does not automatically expose your computer, unrelated accounts, or complete host-chat history; Corvio receives the tool requests and content the host sends. Granted tools can still read or change Corvio data, so review the list before approving. Existing connections keep their prior attribution until you reconnect OAuth.
If an Agent is helping you install
The Agent should explain the benefit for your work, identify the official source, describe what will be installed, and summarize the permissions and removal path. It should ask once if you have not already confirmed.
After you explicitly confirm, it should continue when the source and requested permissions match and the host allows it. It should not loop on the same generic warning. A real source or hash mismatch and a host or organization policy are valid reasons to stop.
Confirm it works and keep control
Ask the Agent to list the Corvio Workspaces you can access, then find one document you recognize. If you approved writing, try one reversible change and read it back. A downloaded ZIP or completed consent page alone does not prove the connection works.
To stop access, remove the Corvio Connector in this host and revoke the Remote MCP connection credential in Corvio Settings → API keys. If the host still asks for authentication after approval, remove the stale connection and begin a new one; old authorization links cannot be reused.
- List authorized Workspaces.
- Find and open one known item.
- Try and read back one reversible write only if needed.
Follow long-running work without losing control
ask_corvio returns one durable operation ID. The Agent can call get_question with a bounded wait, then copy progress_cursor into after_cursor to receive only later user-visible milestones. These updates summarize the current phase; they never expose hidden reasoning or count as a final answer.
To stop the work, the Agent calls cancel_question on that same operation. cancellation_requested means Corvio is still stopping it; only cancelled is terminal, and effects that already settled remain recorded. To correct or redirect the work, finish or cancel the current operation, then ask a successor Question with its conversation_id.
- Keep the original operation ID.
- Read only new milestones with the returned cursor.
- Wait for a terminal Question before claiming completion.
- Cancel explicitly and verify the terminal state before saying it stopped.