claude-devtools
A debugging tool for Claude Code that reads the session logs already saved on disk and shows exact file paths, diffs, model thinking and a category-by-category
Low risk
We rate an entry low when it mostly gives the agent instructions and reference material.
Why this level
- Reads local Claude Code session logs, which can contain conversation and code content
- The standalone Docker mode makes zero outbound network calls according to the README
Install
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/claude-devtoolsDetects the agents on your machine, checks the risk and pins the version.
Other ways to install
Install the tool
brew install claude-devtoolsCommands for macOS and Linux, on Windows run them in Git Bash.
Install on macOS with brew install --cask claude-devtools, or download the build for your OS from the releases page, or run it without Electron: docker compose up and open http://localhost:3456.
Other ways from the author
brew install --cask claude-devtoolsmacOS install.
This is third-party code. Review the repository files before installing.
What it does
claude-devtools reads the session and log files Claude Code already writes to ~/.claude on disk and expands them into a detailed interface: instead of a collapsed Read 3 files it shows exact paths and syntax-highlighted content, instead of Edited 2 files it shows a line-by-line diff. It separately visualizes context usage across seven categories (CLAUDE.md, skills, @-mentioned files, tool I/O, thinking, team overhead, user text), a full subagent call tree with per-agent tokens, duration and cost, and the moment context gets compacted. It can inspect sessions on a remote machine over SSH and renders MEMORY.md files as a linked index. The tool wraps or changes nothing in Claude Code itself, it only reads logs that already exist.
Who it is for. For developers who need more detail than the terminal shows about what Claude Code actually did during a session, especially since the terminal started collapsing output into summaries.
Good fit when
- You need to see exactly which files and lines the agent changed, not just a collapsed summary
- You need to figure out why context got compacted and what was lost in the process
- You need a subagent call tree with per-agent tokens and cost
Not a fit when
- The default terminal output and the --verbose flag are already enough
- You need a tool that changes Claude Code's behavior, not just displays its logs
Example request
Open the latest Claude Code session in claude-devtools and show me where the context went before it got compactedLimitations
It shows only what Claude Code already recorded to disk in logs and transcripts, other agents are not supported. The Electron desktop build covers macOS, Linux and Windows, the standalone mode without Electron needs a separate Docker or Node.js run.
How to disable. Remove the app the normal way for your OS (brew uninstall --cask claude-devtools on macOS) or stop the Docker container.
Security check
- Reads local Claude Code session logs, which can contain conversation and code content
- The standalone Docker mode makes zero outbound network calls according to the README
README in short
The README describes the problem: starting from a certain version, Claude Code began collapsing output into short summaries with no file paths or content. claude-devtools reads the same files under ~/.claude and reconstructs the detail: exact paths and diffs, model thinking content, a seven-category context breakdown, a subagent tree with tokens and cost, compaction visualization, SSH session inspection and MEMORY.md rendering. Install options are Homebrew, a direct download for macOS, Linux and Windows, or Docker without Electron. MIT licensed.
FAQ
Are API keys required?
No, the tool reads Claude Code session files that already exist on disk and needs no keys or configuration.
Does claude-devtools change how Claude Code behaves?
No, the README states it is not a wrapper, it only reads logs that already exist and does not intercept Claude Code itself.
Related
A skills library that gives coding agents a development process: brainstorming, planning, TDD, subagents and code review
Skills for real engineers by Matt Pocock
Skills For Real Engineers
Small composable skills for engineering with agents: plan grilling, TDD, bug diagnosis, code review and architecture
GitHub toolkit for spec-driven development: the specify CLI adds agent commands and skills to a project, from principles to implementation
Reference MCP servers
Model Context Protocol servers
Official reference MCP servers: Filesystem, Fetch, Git, Memory, Sequential Thinking, Time and Everything