Query from an AI agent (MCP)
Give an AI coding agent or assistant the ability to ask about your customers over MCP.
Atune's understanding isn't only for your code — you can hand it to an AI agent.
The atune mcp server exposes understand() over the
Model Context Protocol, so an assistant can answer questions about your customers
grounded in real signals.
Start the server
npx atune mcpThis runs a stdio MCP server with a single understand tool, authenticated with
your ATUNE_API_KEY.
Point your agent at it
Add it to your MCP client's config. For a stdio server the shape is typically:
{"mcpServers": { "atune": { "command": "npx", "args": ["atune", "mcp"], "env": { "ATUNE_API_KEY": "<YOUR_API_KEY>" } }}}Sign in to see this sample with your own workspace API key.
Now your agent can ask things like "which accounts look at risk this week?" or "why did Acme go quiet?" and get a cited answer.
It can't make things up
Answers come back cited, with knows: false when there isn't enough
understanding yet. The agent reports what your signals actually support — it can't
fabricate a customer story.
Why this matters
This is the manifesto's end state: understanding as the layer every AI agent relies on. The same grounded answer your dashboard-free code queries, an agent can query too — on your behalf.