Trace the workflow
Capture prompts, structured claims, model outputs, retrieval steps, and agent tool execution in one inspectable timeline.
AI OBSERVABILITY & EVALUATION
Make AI behavior
visible and testable.
ReasonLens helps you inspect how AI answers are produced—from model requests and retrieved evidence to agent tools, decisions, and evaluation results.
Trace explorerEvidence provenanceRun comparisonsOpen source
R
Trace explorerWorkflow run · completed
● CompletedANSWER TRACE
Should we trust this answer?
Model requestPrompt + generation settings
CompleteEvidence retrieval3 sources · 2 cited
InspectFinal answerClaims linked to evidence
ViewEvidence coverage92%
Trace duration1.42s
Understandwhat happened
Inspectwhy it happened
Comparewhat changed
Evaluatewhether it improved
BUILT FOR ITERATION
From opaque outputs to inspectable runs.
ReasonLens is open source and helps developers inspect recorded AI execution, compare run behavior and evidence, and evaluate observable quality signals—without claiming to reveal hidden model reasoning.
Follow the evidence
Explore retrieval evidence, citations, provenance, and the links between source material and an answer.
Compare changes
Branch from a prior trace, compare variants side by side, and see what changed between runs.
Evaluate regressions
Run evaluation datasets, review quality signals, and use regression gates to spot changes before shipping.
A PRACTICAL WORKFLOW
Capture. Inspect. Improve.
Start with a prompt, choose a trace mode, and review the structure that the workflow records. Revisit past runs, branch experiments, and evaluate results as your implementation evolves.
Open the workspace →1Run a traceCapture a chat, RAG, or agent execution.
2Inspect the detailsFollow evidence, provenance, claims, and tools.
3Compare and evaluateReview variants and test quality changes.