Exports and AI handoff

Needs

Nothing beyond the extension

Produces

report.md, session.json, MANIFEST.md, transcript.json, assets/

Optional

report.html and openapi.json

What is in the ZIP

• report.md: the timeline as Markdown, with [mm:ss] timestamps joining interactions, requests, console output, annotations and narration.
• session.json: the structured events for scripts and tools.
• MANIFEST.md and transcript.json: what is included and the narration text.
• assets/: screenshots, request bodies, uploaded files and, at higher levels, video and audio.
• report.html (optional): a standalone page with screenshots embedded, for someone who just wants to read it.
• openapi.json (optional): an API description inferred from observed JSON traffic when API spec capture is on.

Detail levels

Long sessions are large. Pick a level and the export panel shows an estimated token count for each, so the report fits the context window of the model you are using. Trimming is deterministic and makes no model call, and the same recording can be exported again at a different level.

Handing it to an agent

Unzip the export inside the repository and point the agent at report.md. The guides for Claude Code, Cursor, Codex and ChatGPT include prompts that explain the timestamps. The example report shows the format in full.

Limits

• Exports contain whatever was captured after redaction. Review them before sharing outside your team.
• The standalone HTML report does not embed video or audio.