Can an AI SRE restart or fix things itself when it finds failed deploys?

A Tares agent reads and concludes by default, and you let one agent act by connecting an external MCP server to it. Without a server, the agent investigates failed deploys and reports what it finds, with no restarting, deploying or ticketing. With a server, that one agent can use the tools the server exposes. Every run records the external tools it called.

How it works

Every Tares agent starts with two read tools. Both cover the sources of the projects the run is for, unless the agent names others with sources.

• read returns one correlated timeline for an entity.
• stats returns counts per value of one label, this window against the window before, largest change first. It suits a busy stream where reading lines would only show the latest few.

The agent can also end its runs with the conclude tool. That is the default boundary: the agent reads and concludes, and it does not act.

You move the boundary with external MCP servers. Tools from a server can act on whatever the server exposes, so you decide per agent which servers it may touch. At run time, the selected servers’ tools are offered to the model next to read and stats, prefixed by server name, for example github__create_issue.

Guardrails apply to every external call:

• A tool result is capped at 8,000 characters, so a verbose server cannot crowd out the evidence.
• A server that is down when the run starts is skipped and logged. The run continues with the tools that remain.
• A failed call is returned to the model as an error to reason about.

The Tools and MCP servers page covers the full detail.

Set it up

To let one agent act on what it finds, register a server once and then opt the agent in.

1. Open Settings → MCP servers in a project’s Advanced setup. Any project’s agents can then use a server you register here.
2. Add the server with a name and a streamable-HTTP URL. Stdio servers are not supported.
3. Add an auth header if the server needs one. The value is stored as a secret and never shown again. It can also point at a stored GitHub credential with credential:github/<name>, so rotating the token in one place rotates the server too.
4. Under Advanced: extra headers, add any non-secret headers you want sent with every request. GitHub’s hosted MCP server takes these two to narrow what its tools can do:

X-MCP-Toolsets: repos,pull_requests
X-MCP-Readonly: true

X-MCP-Toolsets limits the tools the server offers. X-MCP-Readonly: true gives you a read-only agent. 5. Select Test. Tares connects live and lists the server’s tools, so you see what the agent could call before it runs. 6. Open the agent’s form and pick the server under External tools. Each selection shows as a removable tag, and removing the tag takes the tools away from that agent again.

Other agents are not affected. Only agents you opt in get the server’s tools. If you want a diagnosis agent for failed deploys that never acts, leave External tools empty on it. Tares agents are covered in AI SRE.

What you get

• A read-only default. An agent that investigates failed deploys reads the correlated timeline, concludes and reports, and nothing outside Tares changes.
• Per-agent control. You choose which agents may touch which servers, and you can narrow a server with extra headers such as X-MCP-Readonly: true.
• An audit trail. Each run records the external tools it called, shown on the run under Runs. You can see what an agent touched outside Tares, run by run.
• Resilient runs. A down server is skipped, a failed call becomes an error the model reasons about, and a tool result never exceeds 8,000 characters.
• A live check before you commit. Test lists the server’s tools, so you know what you are enabling.

When something else fits

• If you only want to know why a deploy failed and who should look, keep the default. The agent reads, concludes and reports, and you or your own automation take the action.
• If you want a person to approve every restart or fix, leave External tools empty and act on the conclusion yourself, or give the agent a read-only server with X-MCP-Readonly: true.
• Tares does not run restarts or deploys on its own. Any action comes from tools on a server you register, so your server’s own permissions set how far the agent can go.

Related