parse-cowork-session

Parsing a cowork session JSONL file involves filtering out session metadata, attachment records, and tool-response payloads to isolate user and assistant turns. A bundled Python parser processes the chat log and optionally writes the cleaned conversation as a markdown output file for documentation.

The best way to clean chat logs is to use a parsing script that skips session metadata, attachment records, and tool-response payloads. By filtering this noise, the parser isolates the core user and assistant turns, producing a polished markdown transcript that is easy to share and review.

Yes, the transcript parser processes UUID-named JSONL files specifically from claude-cowork session exports. It reads the chat log structure, filters out tool-response payloads and metadata, and extracts the preserved user and assistant turns for markdown rendering.

Yes, you can print the parsed chat log inline for quick inspection instead of writing a markdown output file. The parser offers optional file output, letting you choose between saving the rendered conversation as markdown or displaying the cleaned transcript directly in your console.

When rendering a cowork session transcript, the parser skips session metadata, attachment records, and tool-response payloads. Filtering these records from the JSONL chat log ensures the markdown output preserves only the meaningful user and assistant turns, keeping the final transcript clean and readable.