# Claude DevTools Published January 1, 2026 · Updated August 10, 2026 Visualise and analyse Claude Code session executions Category: Developer Tools · Free · Replaces LangSmith ($39/month) · [Official website](https://github.com/matt1398/claude-devtools) ## Install with Homebrew `brew install --cask claude-devtools` ## Quick take Claude DevTools is the missing X-ray for Claude Code on Mac: free, local, and focused on tool traces and tokens. Install it if you run agents daily or debug cost spikes; skip it if Claude Code is an occasional toy. Keep expectations clear: it observes and explains; it does not replace Anthropic's agent. Best for: Daily Claude Code users optimising prompts and cost, Leads reviewing agent behaviour, Security-conscious devs watching sensitive path access, Mac users with remote Linux devboxes via SSH ## What is Claude DevTools? Claude DevTools (also styled claude-devtools) is an open-source desktop inspector for Claude Code sessions. It does not wrap or replace Claude Code. Instead it reads the session logs already stored under ~/.claude/ on your machine and reconstructs what the agent actually did: thinking steps, tool calls, diffs, subagent trees, token attribution, and project memory files. Why it exists: as Claude Code matured, terminal UI sometimes collapsed rich traces into opaque summaries. Community backlash against "coding blind" experiences led to tools that restore visibility without forcing --verbose JSON spam. As of August 2026 the project markets itself to 50,000+ developers, ships macOS (Apple Silicon and Intel), Windows, and Linux builds, and installs on Mac via `brew install --cask claude-devtools`. It is MIT licensed, requires no API keys, and makes zero outbound calls for core inspection. That privacy posture suits developers who dislike SaaS log uploads. Pricing is free. There is no premium tier. You already pay Anthropic for Claude Code usage; DevTools only visualises local artifacts. Relationship to other tools: complementary to Vibe Notch / Claude Island (live status) rather than competing. DevTools is forensic and analytical after or during sessions; notch apps are real-time HUDs. It is not an official Anthropic product, though it is tightly coupled to Claude Code log formats and must track upstream changes. On Mac, the app is a practical companion for anyone on Pro/Max who wants to understand token burn, debug why an agent went sideways, or compare two prompt strategies side by side. If you only run Claude Code once a month, you may not need it. If Claude Code is daily infrastructure, DevTools quickly pays for itself in saved confusion, at $0 software cost. For teams, Claude DevTools becomes a coaching instrument: leads can open a junior's session (with consent and local access) and show exactly where the agent reread the same failing test fifty times. That teaching moment is hard to recreate from a final git diff alone. In consulting settings, DevTools also helps justify invoices. You can show that a day of agent work touched the right services rather than thrashing unrelated packages. Technical caveats worth repeating: because the tool depends on on-disk log schemas, a Claude Code upgrade can temporarily break parsing until DevTools ships a matching release. Keep both tools reasonably current. If you encrypt home directories or use unusual Claude config locations, point the app at the correct paths. On managed Macs, ensure endpoint security tools do not quarantine the app when it reads many small session files quickly. Feature depth in practice: context reconstruction charts often reveal that half of a '200k context' feeling is tool output residue and repeated file reads, not your carefully written prompts. Cleaning CLAUDE.md and adding ignore patterns can cut costs more than switching from Opus to Sonnet. Subagent trees similarly show whether parallelisation helped or merely multiplied confusion. Notification triggers on.env access are the closest thing to a runtime seatbelt DevTools offers, configure them before the first untrusted repo experiment. Comparison with simply using jq on JSONL logs: scripts are fine for one-off forensics, but DevTools wins for daily navigation, multi-pane compare, memory browsing, and SSH remote hosts. Comparison with IDE local history: IDEs show code evolution, not why the agent chose a tool sequence. You want both. Comparison with Anthropic's official UIs: official products optimise for doing work; DevTools optimises for understanding work already done. In short, if Claude Code is part of your daily Mac craft, Claude DevTools is the observability pane you will eventually install anyway, better to install it before the first mysterious Max invoice or the first baffling wrong refactor you cannot explain to a teammate. ## Features - **Session transcript browser.** Open any historical Claude Code session from disk and scroll human-readable messages instead of raw JSONL. - **Tool call inspector.** Expand Read/Edit/Bash/Grep calls with syntax highlighting and inline diffs, the opposite of a collapsed "Read 3 files" summary. - **Context reconstruction.** Per-turn token attribution across categories such as CLAUDE.md, skills, @-files, tool I/O, thinking, and user text so you see what filled the window before compaction. - **Subagent and team trees.** Visualise nested agents with their own token, duration, and cost metrics. - **Project memory pane.** Browse ~/.claude/projects/./memory layers with markdown rendering and open-in-editor hooks for Cursor, VS Code, Zed, Xcode, and popular terminals. - **Notifications and triggers.** Alerts for.env access, tool errors, high token use, and custom regex on sensitive paths. - **SSH remote sessions.** Inspect ~/.claude on remote machines over SSH using your local UI. Handy for cloud devboxes from a Mac laptop. - **Multi-pane compare.** Open two sessions side by side to A/B prompt strategies with real history rather than vibes. - **Command palette search.** Cmd+K style search across sessions with highlighted snippets to jump to exact messages. - **Copy-friendly rendering.** Selectable text without terminal ANSI pain, copy paths, code blocks, and responses cleanly on Mac. ## How to Install Claude DevTools on Mac Homebrew cask is the path of least resistance on macOS. 1. **Install via Homebrew.** `brew install --cask claude-devtools` 2. **Launch and grant folder access if prompted.** The app needs to read ~/.claude/ logs. 3. **Open a past session.** Pick a project and session; verify tool calls render. 4. **Optional: configure notification triggers.** Add regexes for sensitive paths before your next long agent run. ## Pros - Restores visibility Claude Code terminal summaries remove - Free, local-first, no API keys - Excellent token and subagent forensics for cost control - SSH remote inspection from Mac - Homebrew cask makes updates easy - Complements notch status tools without overlapping ## Cons - Unofficial, can break when Claude Code log formats change - Does not prevent bad agent actions in real time (mostly inspects) - Another Electron-or-desktop app in the dock for minimalists - Requires existing ~/.claude history to be useful - Not a substitute for Anthropic's official product support ## Closer look: Observability for coding agents Agentic coding without traces is ops without metrics. You can always grep JSONL with jq for a one-off autopsy, but day-to-day work is faster in DevTools: multi-pane compare, memory browsing, and SSH remote hosts without reinventing the viewer each time. IDE local history answers what the code looked like over time; it does not explain why the agent picked a tool sequence. Keep both. Anthropic's own products are built to get work done; this inspector is built to understand work that already happened. ## FAQ ### Is Claude DevTools official from Anthropic? No. It is an independent open-source project that reads local Claude Code logs. ### Does it upload my code? Core inspection is local and marketed as zero outbound calls. Still review the app's permissions and release notes for any optional analytics. ### How do I install on Apple Silicon? Use `brew install --cask claude-devtools` or the Apple Silicon.dmg from GitHub releases. ### Can it inspect sessions from last month? Yes, any sessions still on disk under ~/.claude/. ### Does it work without Claude Code Pro? It works on whatever sessions exist. Without Claude Code usage you will have nothing to inspect. ### Is it free for commercial use? MIT licensing generally allows commercial use; verify the LICENSE file in the repo for your compliance process. ### Does Claude DevTools work on Linux servers? Yes, native Linux builds exist, and macOS clients can inspect remote Linux ~/.claude directories over SSH. ### Will it slow down Claude Code? No, it reads logs; it is not in the generation path. Impact is disk reads when you open the UI. ### Can I export reports for finance? Use the UI's token breakdowns and your own screenshots/CSV workflows; treat it as an engineering observability tool first, not a formal billing system of record. ### Should every Claude Code user install this? Install when you multitask away from the terminal regularly or need faster permission UX. Skip if you always watch the agent full-screen on a second display and dislike extra menu bar apps. ## Sources - [claude-devtools official site](https://claude-dev.tools/) - [claude-devtools GitHub](https://github.com/matt1398/claude-devtools) - [Claude Code product](https://www.anthropic.com/claude-code) - [Claude pricing (usage context)](https://claude.com/pricing) - [Homebrew cask install line](https://claude-dev.tools/) - [Claude Code ecosystem companions discussion context](https://news.ycombinator.com/item?id=46978710) ## Related - [Cursor](https://bundl.run/apps/cursor) - [Claude Code](https://bundl.run/apps/claude-code) - [ChatGPT](https://bundl.run/apps/chatgpt) - [Claude](https://bundl.run/apps/claude) - [Codex](https://bundl.run/apps/codex) - [Windsurf](https://bundl.run/apps/windsurf) - [Claude DevTools vs CodexMonitor](https://bundl.run/compare/claude-devtools-vs-codexmonitor) - [Claude DevTools vs supacode](https://bundl.run/compare/supacode-vs-claude-devtools) - [Free alternative to LangSmith](https://github.com/matt1398) ```json { "@context": "https://schema.org", "@graph": [ { "@type": "Organization", "@id": "https://bundl.run/#organization", "name": "Bundl.run", "url": "https://bundl.run", "logo": { "@type": "ImageObject", "url": "https://bundl.run/og-image.png", "width": 1200, "height": 630 }, "description": "The Ninite for Mac. Install all your essential Mac apps with one terminal command.", "sameAs": [ "https://github.com/abhiofficial/bundl-mac-setup", "https://x.com/bundlrun", "https://www.producthunt.com/products/bundl-run" ], "foundingDate": "2024", "contactPoint": { "@type": "ContactPoint", "contactType": "customer support", "url": "https://bundl.run/faq" } }, { "@type": "WebSite", "@id": "https://bundl.run/#website", "name": "Bundl.run", "url": "https://bundl.run", "description": "The Ninite for Mac. Install all your essential Mac apps with one terminal command.", "publisher": { "@id": "https://bundl.run/#organization" }, "inLanguage": "en-US" }, { "@type": "Person", "@id": "https://bundl.run/authors/alex-chen#person", "name": "Alex Chen", "jobTitle": "Senior Developer Tools Specialist", "url": "https://bundl.run/authors/alex-chen", "worksFor": { "@id": "https://bundl.run/#organization" }, "description": "Alex Chen has been evaluating developer tools and productivity software for over 12 years, with deep expertise in code editors, terminal emulators, and development environments. As a former software engineer at several Bay Area startups, Alex brings hands-on experience with the real-world workflows these tools are meant to enhance. Alex tests each application extensively on both Intel and Apple Silicon Macs, documenting performance metrics, integration capabilities, and workflow efficiency. When not reviewing software, Alex contributes to open-source projects and writes technical tutorials for the developer community.", "knowsAbout": [ "Code Editors & IDEs", "Terminal Emulators", "Version Control Tools", "DevOps & CI/CD", "API Development", "Performance Benchmarking" ], "image": "https://bundl.run/authors/alex-chen.svg" }, { "@type": "BreadcrumbList", "@id": "https://bundl.run/apps/claude-devtools#breadcrumb", "itemListElement": [ { "@type": "ListItem", "position": 1, "name": "Home", "item": "https://bundl.run" }, { "@type": "ListItem", "position": 2, "name": "Apps", "item": "https://bundl.run/apps" }, { "@type": "ListItem", "position": 3, "name": "Claude DevTools", "item": "https://bundl.run/apps/claude-devtools" } ] }, { "@type": "WebPage", "@id": "https://bundl.run/apps/claude-devtools", "url": "https://bundl.run/apps/claude-devtools", "name": "Claude DevTools for Mac", "description": "Visualise and analyse Claude Code session executions", "isPartOf": { "@id": "https://bundl.run/#website" }, "publisher": { "@id": "https://bundl.run/#organization" }, "inLanguage": "en-US", "datePublished": "2026-01-01T00:00:00Z", "dateModified": "2026-08-10T13:34:04.000Z", "author": { "@id": "https://bundl.run/authors/alex-chen#person" }, "mainEntity": { "@id": "https://bundl.run/apps/claude-devtools#software" }, "breadcrumb": { "@id": "https://bundl.run/apps/claude-devtools#breadcrumb" } }, { "@type": "SoftwareApplication", "@id": "https://bundl.run/apps/claude-devtools#software", "name": "Claude DevTools", "description": "Visualise and analyse Claude Code session executions", "applicationCategory": "DeveloperApplication", "operatingSystem": "macOS", "url": "https://bundl.run/apps/claude-devtools", "image": { "@type": "ImageObject", "url": "https://img.logo.dev/anthropic.com", "caption": "Claude DevTools app icon for Mac" }, "sameAs": [ "https://github.com/matt1398/claude-devtools", "https://formulae.brew.sh/cask/claude-devtools" ], "review": { "@type": "Review", "author": { "@id": "https://bundl.run/authors/alex-chen#person" }, "reviewBody": "Claude DevTools is the missing X-ray for Claude Code on Mac: free, local, and focused on tool traces and tokens. Install it if you run agents daily or debug cost spikes; skip it if Claude Code is an occasional toy. Keep expectations clear: it observes and explains; it does not replace Anthropic's agent." }, "about": { "@type": "Thing", "name": "Claude DevTools", "description": "Visualise and analyse Claude Code session executions" }, "isPartOf": { "@id": "https://bundl.run/#website" } }, { "@type": "Article", "@id": "https://bundl.run/apps/claude-devtools#article", "headline": "Claude DevTools for Mac — Full Review & Installation Guide 2026", "description": "Claude DevTools is the missing X-ray for Claude Code on Mac: free, local, and focused on tool traces and tokens. Install it if you run agents daily or debug cost spikes; skip it if Claude Code is an occasional toy. Keep expectations clear: it observes and explains; it does not replace Anthropic's agent.", "image": "https://img.logo.dev/anthropic.com", "author": { "@id": "https://bundl.run/authors/alex-chen#person" }, "publisher": { "@id": "https://bundl.run/#organization" }, "datePublished": "2026-01-01T00:00:00Z", "dateModified": "2026-08-10T13:34:04.000Z", "mainEntityOfPage": { "@type": "WebPage", "@id": "https://bundl.run/apps/claude-devtools" }, "articleSection": "DeveloperApplication", "speakable": { "@type": "SpeakableSpecification", "cssSelector": [ "h1", ".key-facts" ] }, "about": { "@type": "SoftwareApplication", "name": "Claude DevTools", "url": "https://bundl.run/apps/claude-devtools" }, "mentions": [ { "@type": "SoftwareApplication", "name": "LangSmith", "url": "https://github.com/matt1398" } ] }, { "@type": "FAQPage", "@id": "https://bundl.run/apps/claude-devtools#faq", "mainEntity": [ { "@type": "Question", "name": "Is Claude DevTools official from Anthropic?", "acceptedAnswer": { "@type": "Answer", "text": "No. It is an independent open-source project that reads local Claude Code logs." } }, { "@type": "Question", "name": "Does it upload my code?", "acceptedAnswer": { "@type": "Answer", "text": "Core inspection is local and marketed as zero outbound calls. Still review the app's permissions and release notes for any optional analytics." } }, { "@type": "Question", "name": "How do I install on Apple Silicon?", "acceptedAnswer": { "@type": "Answer", "text": "Use `brew install --cask claude-devtools` or the Apple Silicon.dmg from GitHub releases." } }, { "@type": "Question", "name": "Can it inspect sessions from last month?", "acceptedAnswer": { "@type": "Answer", "text": "Yes, any sessions still on disk under ~/.claude/." } }, { "@type": "Question", "name": "Does it work without Claude Code Pro?", "acceptedAnswer": { "@type": "Answer", "text": "It works on whatever sessions exist. Without Claude Code usage you will have nothing to inspect." } }, { "@type": "Question", "name": "Is it free for commercial use?", "acceptedAnswer": { "@type": "Answer", "text": "MIT licensing generally allows commercial use; verify the LICENSE file in the repo for your compliance process." } }, { "@type": "Question", "name": "Does Claude DevTools work on Linux servers?", "acceptedAnswer": { "@type": "Answer", "text": "Yes, native Linux builds exist, and macOS clients can inspect remote Linux ~/.claude directories over SSH." } }, { "@type": "Question", "name": "Will it slow down Claude Code?", "acceptedAnswer": { "@type": "Answer", "text": "No, it reads logs; it is not in the generation path. Impact is disk reads when you open the UI." } }, { "@type": "Question", "name": "Can I export reports for finance?", "acceptedAnswer": { "@type": "Answer", "text": "Use the UI's token breakdowns and your own screenshots/CSV workflows; treat it as an engineering observability tool first, not a formal billing system of record." } }, { "@type": "Question", "name": "Should every Claude Code user install this?", "acceptedAnswer": { "@type": "Answer", "text": "Install when you multitask away from the terminal regularly or need faster permission UX. Skip if you always watch the agent full-screen on a second display and dislike extra menu bar apps." } } ], "isPartOf": { "@id": "https://bundl.run/apps/claude-devtools" } } ] } ```