cjm-harness-transcripts 0.0.1
A new Python library, harness-transcript, is available for parsing agent-harness session logs. It converts Claude Code JSONL transcripts into a directed acyclic graph, enabling analysis of conversational flows and extraction of user-facing text.
Key takeaways
- Parses agent-harness session transcripts.
- Converts logs into a directed acyclic graph.
- Extracts user-facing text from conversations.
- Supports Claude Code JSONL format.
Why it matters
This tool helps AI developers and researchers better understand and analyze the outputs of AI agent interactions. It allows for structured examination of conversations, improving the development and refinement of AI assistants.




