Claude DevTools is a free, open-source Mac utility that turns raw Claude Code session data into interactive visual timelines and usage breakdowns, giving developers a clear audit trail of every AI-assisted coding run.
What is Claude DevTools?
Claude DevTools is a desktop companion for Claude Code power users — it reads your local session execution logs and renders them as navigable visual reports, so you can see exactly what happened during any AI coding session: which tools fired, how long each step took, what tokens were consumed, and where things went sideways.
If you have ever squinted at Claude Code's raw JSONL transcripts trying to reconstruct why a long agentic run stalled or burned through your context budget, this tool is the answer. It surfaces that data in a form you can actually reason about.
What does Claude DevTools do best?
Claude DevTools excels at turning opaque session logs into a scannable execution timeline. Each tool call — file reads, shell commands, code edits — appears as a discrete node you can inspect individually, complete with timing information and input/output payloads.
- Session timeline view: chronological breakdown of every step in a Claude Code run, colour-coded by tool type
- Token usage analysis: see where context budget was spent across the session, which is invaluable when debugging runaway agents
- Tool-call inspection: drill into any individual call to see its arguments and response without parsing raw JSON by hand
- Error surfacing: failed steps and retries are visually distinct, making post-mortem analysis quick
I have found it particularly useful after long autonomous runs — the kind where Claude Code rewrites half a codebase over 40+ steps. Scrolling back through the transcript to understand a decision is painful; the visual timeline makes it a two-minute scan instead.
Is Claude DevTools free?
Yes — Claude DevTools is completely free to download and use. It is an open-source project hosted on GitHub, so you can inspect the source, file issues, or contribute a pull request. There is no paid tier, no usage cap, and no account required. The only dependency is having Claude Code itself installed and having at least one recorded session on disk.
Who should use Claude DevTools?
Claude DevTools is squarely aimed at developers who use Claude Code regularly and want more observability into what their AI assistant is actually doing. If you run quick one-off prompts and rarely revisit them, the overhead of opening a separate visualiser probably isn't worth it. But if you:
- Run long multi-step agentic sessions (refactors, migrations, bulk code generation)
- Are hitting token budget limits and want to understand why
- Debug unexpected Claude Code behaviour after the fact
- Need an audit trail of AI-assisted changes for a team or client
…then Claude DevTools fills a gap that nothing else currently covers. Tools like the Claude.ai web console show you a conversation view, not an execution trace. Claude DevTools is explicitly about the execution layer.
What are the best Claude DevTools alternatives?
There is no direct equivalent right now — that is part of what makes this tool interesting. For general AI session inspection, your closest options are reading Claude Code's raw session JSONL files in a text editor, or using the built-in --verbose flag during a run. Neither gives you the interactive visual breakdown Claude DevTools provides.
For broader LLM observability (tracing, cost tracking, multi-provider support), tools like LangSmith, Langfuse, and Helicone cover different ground — they are platform-level monitors, not local session analysers tied to a specific agent tool. If you are working outside Claude Code, those are worth a look. Inside Claude Code's ecosystem, Claude DevTools is currently the only dedicated session inspector I know of.
How actively is Claude DevTools maintained?
Claude DevTools is an early-stage open-source project — feature velocity is tied to community interest and the author's time. As Claude Code itself evolves rapidly (the session format and tool set have changed with recent releases), there is inherent risk that a log format update breaks the parser. Check the GitHub repository's issues and commit history before adopting it in a production workflow. That said, the code is open — if something breaks, you can fix it, and the community seems engaged. For a free tool solving a genuinely underserved problem, the risk-reward is sensible.
