01 / 03run 284 · step 08
Trace the intent
Your framework records the model decision, tool call, and expected result.
See why a browser-agent run went wrong
Connect replay to the trace and eval for every browser-agent run. When an agent stalls or returns the wrong result, open what it saw at the exact action where the browser state diverged.
agent run #284 · book the earliest train
tool trace
01inspectDOM
02click#search
03waitresults
04select08:42
what the agent saw
travel.example.com/searchrec
08:42 → 10:15€42
09:10 → 10:48€38
Continueses_agent_284
47 events · one clock
Browser automation products built with rrweb.
Open the browser state behind any tool call
A tool trace records the requested action. Replay adds the page state immediately before and after it, including the selected value, validation message, overlay, or disabled control that changed the outcome.
- Open the browser state immediately before and after the action named in the trace.
- Compare the interface the agent expected with the DOM it actually received.
- Mark the first visible divergence so later failures do not hide the original cause.
run comparison · step 08
The interface diverged here
agent expected
/checkout · step 08
fare€42
Continuebrowser showed
/checkout · step 08
fare€57
Fare changed: reselect seatsContinuefirst divergence at 00:12 · replay opens here
Review a failed run without paying to run it again
Attach the recording to the run while it happens. When an eval fails, the reviewer can inspect the final browser state, scrub backward through the actions that produced it, and grade the visible outcome using evidence already stored with the run.
- Grade the visible confirmation, record, or state change captured in the final frame.
- Sample low-scoring runs from production and open the evidence already stored with them.
- Retain useful failures with their release, browser, task, and evaluation result.
eval · run 284
Score the visible outcome
final frame · 00:31
Validation error: booking not confirmedgoal "booking confirmed"
observed "validation error"
score0.25
replay attached to failed eval · watch without rerunning
Build a library of real browser failure modes
Label important actions with a custom event, then group recordings by failure class, site, model, release, or evaluator result. Reviewers can find a useful example quickly and open it at the action that defines the failure.
- Open each example at the action or browser event that defines the failure class.
- Keep prompts and private tool payloads in the trace while replay holds the observable page state.
- Export reviewed sequences for offline evaluation or approved training-data work.
trajectory · ses_agent_284
Intent and effect on one clock
trace
model.calltool.click
browser
dom.mutationtask.failed
0.0s1.4s2.1s
shared run id + timestamps join the trace to the recording
Debug failed runs without running them again
Replay turns a failed automation run into an inspectable browser record. Review production failures, compare eval results, and keep useful examples for later analysis without paying for another execution.
- Open the browser evidence from a failed step before starting another run.
- Keep recordings, review notes, and evaluation results together for repeated failures.
- Add browser evidence to an automation stack without building a recorder and replayer.
product settings · feature packaging
Browser run replay
Customer-facing capability
- Available on
- Pro
- Lives in
- Run details
- Sold as
- Usage-based add-on
- Powered by
- rrweb replay
Customer outcome
Customers can explain a failed agent run without rerunning it.
Keep the trace and replay on the same run clock
Give the trace and rrweb event stream the same run id. Store timestamps around model and tool calls, then open replay from a trace row, evaluation result, or failed step. Playback starts at the corresponding browser action inside the run review UI.
02 / 03ses_agent_284
Record the effect
rrweb captures the DOM state before the action and the interface that followed.
03 / 03open at step 08
Open one review
The run page aligns trace, evaluation result, and replay on the same clock.
Choose the backend that fits your product
Use rrweb Cloud for managed browser recordings, or run the open-source replay stack alongside your traces and evaluations.
Start with rrweb Cloud
Use managed capture, storage, and playback while your system keeps the agent trace, evaluation workflow, and run history.
rrweb Cloud uses 80% less bandwidth and storage.
Build on open source
Run recorder, storage, correlation, and replay inside the same boundary as your agent traces and eval data. You control retention, access, and the run review interface.
Frequently asked questions
How do I record automated browser sessions?
Start the rrweb recorder when the automation run opens the page, assign the recording the same run id as the agent trace, and emit custom events around model or tool calls. Send the resulting browser event stream to your own storage or rrweb Cloud.
What does DOM recording preserve that video leaves out?
DOM recording preserves structured page state and event-level timestamps, which makes a run searchable, segmentable, and easier to align with a trace. Video covers pixels from some canvas and cross-origin content. That broader pixel coverage comes without DOM structure.
What does rrweb add to an agent trace?
An agent trace records prompts, model responses, and tool calls. rrweb records the browser state around those calls, including the DOM the agent acted on and the visible result. Shared timestamps and a run id connect the two.
Can replay make an agent run deterministic?
Replay deterministically reconstructs the captured interface sequence. It does not re-execute network requests, application code, or agent decisions. Use a deterministic test harness when you need to run those systems again.
Can I watch an automated browser session live?
Yes. Live mode sends incremental events to the replayer while the run is active. It stays a few seconds behind the browser and helps an operator inspect a run without taking control from the automation framework.
Can recordings become training or eval data?
Yes, after you apply the consent, masking, access, and retention rules required for that use. The open event format can be exported and segmented around outcomes or custom events.
Why rrweb over a packaged agent-observability SaaS?
rrweb is infrastructure for adding browser replay to automation runs. You get the open recorder, event format, and replayer, while your automation stack keeps its traces, evals, storage, and review workflow. A packaged platform is a finished application with its own interface, data model, and pricing.