Use it from your coding agent
Your coding agent can ask the AI SRE what is broken while you fix it. Connect it to Tares over MCP.
On Tares Cloud the endpoint is your workspace’s address plus /mcp; on your local machine it is the
tares mcp endpoint:
claude mcp add --transport http tares http://localhost:8788/mcpConnecting agents has Cursor, Codex and other clients, and keys.
The tools it sees first
| Tool | What it answers |
|---|---|
what_is_broken | What is wrong right now: the headline, whether anybody is watching, and every open incident with what is failing, the likely cause and the next step. |
how_is_service(service) | One service: its state, what it sends, who watches it, its incidents and its blind spots. |
what_happened(service, around, window) | What happened in a stretch of time: log changes, deploys, alerts, checks, incidents and metric steps. Pass a moment as around to see what led up to it. |
mark_incident(incident, state) | Take an incident (on_it), resolve it, or open it again. |
read keeps working too, for one correlated read of any service’s timeline across every source.
Try:
Use tares: what is broken right now, and what changed on checkout in the hour before it started?Last updated on