Skip to content
GitHub
Start a 14-day trial

Connect with a coding agent

Your coding agent already knows how to wire OpenTelemetry. Give it your endpoint and token, and it does the setup in your repository while you watch. Once your telemetry is in, the same agent takes my diagnosis of a request and works on the fix. This guide runs both legs: agent to atrim.ai, and back.

  1. Open Settings → App Onboarding. Under Admin & Governance, open Settings, then App Onboarding.

  2. Generate a token. Press Generate token, keep Send telemetry as the purpose, and press Generate token in the dialog, then Done. The token fills the prompt in the next step for as long as you stay on this page. Leave, and you generate another.

  3. Open the prompt. Under 2. Send telemetry from your app, choose Have my coding agent do it. Optionally set the Service name to how this app should appear; leave it empty and your agent names each service after what it does. The prompt works in any language: it asks your agent to look before it edits, add atrim.ai as an exporter beside any you already run (or ask you first when the only exporter is set by environment variables), and add OpenTelemetry first if the project has none. Edit it in place if your project needs something said.

  4. Press Copy prompt. The copy carries your full token, which the screen keeps masked.

  5. Paste it into Claude Code, Cursor or any coding agent, opened in your app’s repository. The prompt asks it to start the app, confirm the export is accepted, leave its changes uncommitted and report what it changed, and points it at llms.txt, these docs in a form it can read.

The token can only send telemetry; it can’t read or query your data. Treat it as a credential all the same: ask your agent to keep it in an environment variable or your secret store, never in a committed file. To swap one out, rotate or revoke it under App Onboarding.

The prompt, filled with placeholders, is also on Connect your app if you want to read it before you start.

  1. Make a request to your app, so it produces a trace.

  2. Press I’ve started my app — check now under 3. Check that your telemetry arrived. It shows your telemetry Accepted, then Searchable with the services it saw, then Analysis ready.

  3. Press Open your trace. Your request opens as a waterfall.

If nothing arrives, work down the troubleshooting list.

Now the return trip. Every investigation view in atrim.ai has the same Copy for agent button, and it copies what you’re looking at as text your agent can act on.

  1. Press Analyze in the trace’s header. I read every span, show where the time went and, when the request failed, name the service the failure started in.

  2. Press Copy for agent on the analysis’s Summary card. The confirmation reads Analysis copied — paste into your agent.

  3. Paste it into the same agent session. It receives my findings, the root cause with its evidence and the spans behind it, and an ask to confirm the diagnosis and propose the fix.

For a quick look without an analysis, Copy for agent in the trace’s header, beside Analyze, copies the trace itself and asks where the time went.