Integration

Claude Code session files: what they contain

User-selected local JSONL transcripts or telemetry

Supported data

Tool, token, and agent metrics are strongest when telemetry spans are present.

Privacy

Raw prompts, responses, attachments, local paths, and source files stay on your device in either privacy mode. The only thing stats-with-titles adds is the conversation title.

Limitations

Provider export formats may change. Unknown fields are preserved as warnings instead of silently invented metrics.

Duplicate prevention

Re-imports use stable provider session identity and content normalization hashes. Container file checksums never define session identity.

Metric quality

Unavailable values remain null. Estimates are labelled estimated and exact values are never fabricated.

Troubleshooting

If detection fails, use the preview warnings and confirm the export source/version before importing again.

What a Claude Code export can tell you

Prompts: Available; Counted from user events in the transcript.
Responses: Available; Counted from assistant events. Tool results are results, not new answers, and are not counted again.
Tool calls: Available; Counted from tool invocation events, de-duplicated by call id so a retry is not counted twice.
Agents launched: Available; Counted from subagent start events. A stop event closes a launch rather than opening a new one.
Tokens: Sometimes; Read from the usage block on assistant events. Input excludes cache reads for this source. A transcript recorded without usage leaves tokens unavailable instead of zero.
API-equivalent cost: Sometimes; Estimated from token counts against a dated public price list, and only when both the tokens and a known model rate are present.
Models: Available; Read from the model recorded on each event, so a session that switched models lists every one of them.
Summed session span: Available; First to last event instant in the session.
AI active time: Sometimes; Read from reported turn durations. A transcript that records no turn duration leaves it unavailable.
Estimated interaction time: Shown as AI active time; Not computed for Claude Code, which reports real turn durations. The gap estimate exists for sources that cannot measure the real thing.
User active time: Not in export; No export records how long you spent reading or typing. It could only be inferred by subtracting one estimate from another, and an inferred number shown as a measurement would be a fabrication.
Waiting time: Shown as AI active time; Waiting on the model is exactly the AI active time already shown. Repeating it under a second name would add a tile, not a measurement.
Project grouping: Available; Grouped by the Git remote or working directory recorded in the session, hashed before it leaves the browser.