How do I let my own Claude Code session investigate a cron job that stopped with a project key?
Tares lets your own Claude Code session investigate a stopped cron job through a project key. You create the key under Settings, Keys in the project, then connect with claude mcp add and an Authorization Bearer header. The key reads that project only and records findings there, so your session works from the project’s sources and skills and writes its conclusion where the Tares agents already look.
How it works
A project key ties one Claude Code session to one Tares project. Once connected, the project is implied, so the tools need no project argument.
- The key reads this project only and records findings in it. Any other call answers 403.
- Claude Code reaches Tares through the MCP endpoint, which
tares mcpserves athttp://127.0.0.1:8788/mcp. - It loads the project’s skills with
list_skillsandget_skill, then usesstatsandreadto see which values moved and what the entity’s timeline shows across the project’s sources. - It calls
list_findingsandproject_timelineto see what the Tares agents already concluded, so it builds on that work instead of repeating it. - When it concludes, it calls
record_findingwith the entity, a markdown finding and a one-word lowercase verdict such asrca.
Every finding carries the name you gave the key. For a stopped cron job, the entity is whatever your project’s sources call it, and the timeline shows what its sources last reported. The full walkthrough is in Connect an outside agent to a project.
Set it up
- Check the prerequisites. You need a running Tares instance with a project that has sources and a trigger, such as the AI SRE demo. Start the instance with auth on so keys are checked. Run
tares mcpso the MCP endpoint is up. On an open instance there are no keys, so skip the key step and name the project in your prompts instead. - Open the project, go to Settings → Keys and choose New key.
- Name the key after the agent that will hold it, such as
claude-code. Copy the secret now, because it is shown once. - Connect Claude Code with the key:
claude mcp add --transport http tares http://localhost:8788/mcp \
--header "Authorization: Bearer <project key>"- Check the connection in a Claude Code session:
Use tares: which project can you work in?
It calls list_projects and names the one project the key belongs to.
6. Ask about the cron job:
Use tares: my nightly cron job stopped. What happened to it in the last 30 minutes? Check the project’s skills first, and record what you conclude.
- Over stdio instead of HTTP, start
tares-mcpwithTARES_AUTH_TOKENset to the project key in its environment.
The same steps work for other MCP clients such as Codex or Cursor. Only the connect command differs, and Connecting agents in the guide points to the details.
What you get
Your Claude Code session investigates inside one project and leaves a record there.
- The finding appears as a card under What the agents found on the project’s Overview, under the key’s name.
- In Advanced setup’s Activity it is a thread of its own, marked as recorded from outside Tares.
- It sits on the entity’s timeline in Explore, so the next Tares agent that reads that entity starts from your conclusion.
- Access stays narrow. The key cannot touch other projects, and you cut the agent off by revoking the key under Settings → Keys.
If you want the session woken by the project’s triggers instead of asking it yourself, an agent that runs on its own, such as a service with a webhook, can call join_project with its URL. Every firing of every trigger in the project, including triggers added later, is then sent to that URL. Revoking the key also removes its subscriptions, while Revoke on the outside agent under Agents stops only the deliveries.
When something else fits
- If you want the investigation to run on its own whenever the cron job’s trigger fires, use the project’s Tares agents, or have an outside agent join the project so triggers wake it.
- If your instance runs open with no auth, there are no keys. Skip the key and name the project in your prompts.
Related
- Connect an outside agent to a project
- How do I let my own Claude Code session investigate error spikes with a project key?
- How do I let my own Claude Code session investigate high latency with a project key?
- How do I let my own Claude Code session investigate failed deploys with a project key?
Still stuck? Get help