Debug AI agents with the full context
Use Lunary to trace model and tool runs, inspect conversations and test prompt changes.
$20 per user / month. Payment required to activate your workspace.
Follow a failed task through its model calls, tools and final response, then test the fix.
Instrument
Connect an SDK integration or existing OpenTelemetry traces.
Debug
Find the failing operation in the parent and child runs.
Improve
Test a prompt change against the failing example.
At a glance
| Engineering question | Evidence to retain |
|---|---|
| Why did this fail? | Inputs, outputs, operations, and errors |
| Who was affected? | Safe user/session context and environment |
| What changed? | Release, prompt version, model, and parameters |
| Is the fix better? | Baseline and candidate on relevant examples |
| Can we trust the view? | Runtime completion and ingestion health checks |
Trace one complete task
Lunary supports agent, tool, LLM and chain runs. Connect them with parentRunId, then verify a successful task and a failure before expanding coverage.
See what the application did
Inspect model input and output alongside actual tool results. Link the run to its conversation and release to separate a bad model decision from a failed execution.
Test the next prompt
Use templates and the playground with the failed example and nearby successes. Keep the variables, model and parameters fixed so the comparison is useful.
Questions & answers
Can I use existing OpenTelemetry traces?
Yes. Verify attribute mapping, parent relationships and supported payloads.
Does a local trace prove production coverage?
No. Also verify streaming, errors, shutdown and cancellation in the deployed runtime.
By Lunary · Documentation reviewed Sep 16, 2026
Sources & methodology (7)
Based on vendor documentation, not an independent benchmark or hands-on rating. Features and plans can change.
Build with Lunary.
Trace, evaluate, and improve your AI applications.
$20 per user / month. Payment required to activate your workspace.