Custom wake word for AI Agents (MCP)
Outspoken runs a remote MCP server. Point any MCP-capable agent at it and it can read our integration guides and, with a personal access token, see your own models and benchmarks. It is read-only: it cannot train, spend credits or download files.
Train your wake wordTest any model in the playground first.
91%of wake word utterances caught, on average, in our benchmark
~45 minfrom typing your phrase to a finished model
Why it works well here
- ✓One remote server, no local install for clients with native HTTP support
- ✓Read-only by design, so there's nothing it can break or spend
- ✓Works with Claude Code, Cursor, Claude Desktop, Windsurf and other MCP clients
How to set it up
- 01Add the MCP server. Point your client at https://outspoken.cloud/api/mcp. Claude Code and Cursor connect directly over HTTP; other clients use the mcp-remote bridge.
- 02Create a personal access token (optional). From your dashboard profile, create a token so your agent can see your own models and recommended thresholds, not just the public guides.
- 03Ask your agent. Try: "Use the Outspoken MCP server to set up my wake word in this project. Check which platform I use and follow its guide."
claude mcp add --transport http outspoken https://outspoken.cloud/api/mcpQuestions
Is the MCP server read-only?
Yes. It can read integration guides and, with a token, your own models and benchmarks. It cannot train models, spend credits or download files.
Which clients are supported?
Any MCP client works. Claude Code and Cursor connect directly over HTTP; Claude Desktop, Windsurf and others connect through the mcp-remote bridge.
Do I need a token?
No. Without a token your agent can still read the public integration guides. A personal access token additionally lets it see your own trained models.