clewe
recording every move

See what your
agents actually did.

cleweis the flight recorder for Claude Code & Codex agents. The local agent tails your transcripts and ships every prompt, thought, tool call and result as a live trace waterfall — so you stop guessing what the model did and start shipping better agents.

trace · refactor-auth · claude-code
+0msUSERrefactor the auth module
+0.1sTHINKplanning the change
+0.9sAGENTI'll read the current code first
+1.1sTOOLRead · auth.ts
+1.3sRESULT320 lines
+1.6sTHINKspotting the stale session check
+2.0sTOOLBash · npm test
+3.4sRESULT3 failing
+3.7sAGENTpatching the token guard
+3.9sTOOLEdit · auth.ts
+4.5sTOOLBash · npm test
+6.0sRESULTall green ✓
useragentthinktoolresult

Trace waterfalls

Every prompt, thought, tool call & result — nested by subagent, on a real timeline.

Compare & catch

Diff two runs side by side. Flag the bad ones. See exactly where it went sideways.

Runs → datasets

Turn agent runs into shareable HuggingFace datasets in a click.

Start watching

Sign in to open your agents’ traces.

or email

No account? Create one — it’s free.

two lines to start recording
$ pip install claude-monitor-sdk
$ claude-monitor agent  # tail ~/.claude, ship traces

Claude Code + Codex CLIzero-dependency SDKdocs