Free AI agent log viewer · no account

See what your AI agent actually did.

Traceplain turns Codex exec JSONL, Claude Code stream JSON, ZAK records, OpenTelemetry, and common JSON tool logs into a human handback: what changed, what evidence supports it, what the agent merely claimed, and where you need to step in.

Review an agent log

Runs in your browser. Imported activity never leaves your device.

HAND BACK / 14:32 REVIEW NEEDED

7actions observed

3files changed

1claim not proven

Observed

The agent changed three files and committed them locally.

Reported, not verified

“All tests pass.” No test execution was present in the activity record.

YOUR NEXT DECISIONReview the auth change before allowing a push.

LOCAL-FIRSTNO ACCOUNT REQUIREDEVIDENCE-LABELLEDZAK-READY

TRY TRACEPLAIN

Turn activity into an answer.

Paste JSON or NDJSON from Codex exec, Claude Code, ZAK, OpenTelemetry, or another tool log. Traceplain recognises these documented event shapes and keeps uncertain facts uncertain.

01Activity record

Drop a log here or choose JSON, NDJSON, or text

OR PASTE

Your handback will appear here

Traceplain labels evidence without pretending a log proves more than it does.

THE DIFFERENCE

Logs describe. Traceplain distinguishes.

Observed

A tool result, exit code, changed target, or other event present in the supplied record.

Governed

A policy decision or authority outcome recorded with a receipt reference.

Prevented

An action the record says was denied or blocked before its intended effect.

Reported

A statement by the agent or runtime that Traceplain cannot independently establish.

Inferred

A cautious interpretation drawn from names or context, clearly marked as inference.

Unknown

Evidence is missing, ambiguous, unsupported, or outside the imported record.

NEED A METHOD? Read the practical guide to reviewing an AI agent activity log →

NEED A HANDOFF? Download the free evidence-labelled AI agent handoff template →

USING CODEX? Capture and review codex exec --json output →

USING CLAUDE CODE? Review stream JSON and session tool calls locally →

USING OTEL? Review an OpenTelemetry GenAI trace locally →

AUTOMATING CI? Generate a bounded Markdown handback in GitHub Actions →

FOUNDING REVIEW KIT

A repeatable method for reviewing agent work.

The public viewer stays free. The Review Kit gives you the complete printable method and reusable handback files immediately after checkout.

Eight-page printable review system
Approval gate and seven-day team trial
Worked example and evidence labels
Markdown template and JSON Schema

Preview one of the eight pages

Preview of the five-pass Review Kit worksheet covering requested, authorised, executed, changed, and verified evidence — Page 2 of 8 · Five-pass review worksheet. The paid kit also includes the approval gate, team trial, worked example, handback, and reusable files.

TRACEPLAIN REVIEW KIT

^A$29

One time · immediate digital delivery through Stripe

Get the Review Kit Checkout collects your name and email for the receipt. Imported agent activity is never sent to Stripe.

NOT READY TO BUY?

Tell us what your agent leaves unclear.

Join the field notes or share the reason you stopped. That evidence decides what Traceplain becomes next.