Regression-test agents on real trajectories
Curate golden datasets from production traffic. Each example pins the whole agent run — every step, tool, and MCP call — forever.
Fluiq/ datasetsExamples
48
checkout-regressions
Trajectories
41
full runs pinned
Batch Runs
6
agentic eval · security
Pinned trajectory · example #12
run score 0.88What you get
Capture
Whole-trajectory examples
Add any run from the trace drawer and Fluiq pins its full trajectory — LLM calls, agent steps, tool and MCP calls, media — independent of trace retention.
Sync
Connect Agents
Link a traced agent to a dataset and every run it has ever made is imported, deduplicated — and future runs keep appending automatically.
Batch
Agentic eval & security runs
Re-run agentic evaluation or the full security scan over every example and get a scored report — the regression gate for prompt and model changes.
Enriched
Live quality signals
Each example carries its run's eval scores, security verdicts, and cost, backfilled automatically as workers finish.
Pin a run once. Evaluate it forever.
- Add to Dataset on a root trace snapshots the entire run, media included
- Batch runs feed pinned trajectories to the same agentic evaluator used on live traffic
- Examples stay evaluable after the source trace ages out of retention
# Link a run to a dataset via the API (or one click in the UI)
requests.post(f"{BASE}/datasets/{ds_id}/examples", headers=H, json={
"input": "Refund my last order",
"expected_output": "Opened refund #4821",
# the whole trajectory is pinned from the run's root trace
"metadata": {"source_trace_id": "0f9c...e21"},
})Part of the Fluiq platform
Compare plansUnlimited traces, always free.
Start on the free tier and turn on each pillar as your pipeline grows. No code changes required.