Introduction
What Tracewell records, how a run becomes a trace and where evals and deploys fit.
LAST UPDATED
Tracewell records every run of your agent as a trace, scores it against evals you write and gates deploys on the result.
The three parts
Runs: a readable trace for every call to your agent
Evals: tests that score a version against your cases
Deploy: gates, canaries and rollbacks driven by those scores
How it fits together
You add the SDK to your agent. Each call becomes a run. Evals replay your cases against any version. A deploy promotes a version only when its evals pass.
Where to go next
Start with the quickstart, then read about spans and attributes.