An agent took an unexpected action.
Your trace has the model and tool responses. Provenance shows which value decided the branch that became the action.
Python execution explainability
Retrace records continuously, at under 1% overhead*, so the evidence exists before anyone asks. When a result is questioned, trace it back to the inputs that produced it.
Illustrative checkout example
Recording · always on · under 1% overhead*
Engineer opens Tuesday's recorded run
On supported workloads. Overhead depends on how much I/O an application performs.
Provenance · illustrative example
Select a result in a recorded run. Provenance follows it back through the operations and inputs that produced it, all the way to what the run actually received.
What the run received
What the code did
Combined
Start here
Items 1–2 at 1.0842, items 3–4 at 1.0388 after a background refresh. Using the initial rate of 1.0842 for all four items would have produced £433.68.
The source says this could happen. Only the recording shows it did.
Illustrative example.The questions behind the investigation
Your trace has the model and tool responses. Provenance shows which value decided the branch that became the action.
The calculation succeeded, but its inputs may have changed since. Provenance traces this result back to the inputs that produced it.
A CI test passes on rerun, or a customer's integration fails and nobody can reproduce it. Replay the original run, not a new attempt.
How it works
Recording runs alongside your Python application in your environment, always on. When a result is questioned, cloud Replay and Provenance explain the recorded run, with VS Code on your machine as the developer interface.
Recorded continuously: external calls, returned values, exceptions and thread ordering.
The external interactions and ordering needed to reconstruct the execution.
Return to the recorded run and step forwards and backwards through it in VS Code.
Select a result to see the inputs, operations and source locations that produced it.
Because recording is always on, nobody needs to predict which run will be questioned. A run can report success and still need explaining later.
Private preview
We record one worker or workflow continuously, then explain the next result your team questions. Every pilot includes Provenance.
Talk to the founders
Tell us about an investigation your team found difficult: an unexpected agent action, a disputed result, or a Python failure that was hard to explain.
We'll discuss what happened, what your existing tools recovered, and what evidence was missing.
Retrace is built in London by Henry Yates and Nathan Matthews. We work with engineering teams to explain the results their Python systems produce.