Claude Watch CLI
Claude Watch draws a localhost map of project logic, shows Claude Code tool calls live and answers questions
- 69 ★ on the repository
- Last commit
Extended access
Why this level
- Hooks change Claude Code settings, and the dashboard receives data about every agent action
- Question search sends the contents of project logic files to the Anthropic API
Install
- Works in
Manual install
node dist/cli.js startAfter `npm install` in the root and in `dashboard`, then `npm run build` in the repository clone.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Hooks change Claude Code settings, and the dashboard receives data about every agent action
- Question search sends the contents of project logic files to the Anthropic API
This is third-party code. Review the repository files before installing.
What it does
Claude Watch is an open-source observability dashboard for Claude Code. It discovers the logic files in a project, system prompts, voice guides, rules, configs and playbooks, and shows them as an interactive bubble map you can zoom. A classifier reads file names and content and spots words such as NEVER, ALWAYS, MUST and PRIORITY. `claude-watch init` adds hooks to `.claude/settings.json`, and every Claude Code tool call lands in SQLite and flows over WebSocket to the dashboard at `http://localhost:3853`. That shows what the agent is doing right now. There are project state snapshots (`claude-watch snapshot <name>`) that can be saved and restored, a `start` command for the dashboard and `status` for server stats.
Who it is for. People who build projects around rules, prompts and playbooks for Claude Code and want to see which files shape the agent and what it is doing.
Good fit when
- A project has dozens of rule and prompt files and you want them on one map
- You want to watch which tools Claude Code calls while it works
- Project state is saved as a snapshot before big edits
- A question such as where the testing rule lives is asked in plain words across logic files
- Hooks for all projects are installed once with `init --global`
Not a fit when
- You need ongoing team monitoring: the tool is local and single user
- You do not use Claude Code: the hooks go into its settings
- Project files must not go to an external API: question search uses an Anthropic key
Example request
Start the claude-watch dashboard, attach hooks in this project and show the map of rule filesLimitations
Install runs from a repository clone: `npm install` in the root and in `dashboard`, then `npm run build` and `node dist/cli.js start`. The README names no registry package. Hooks are written to the project `.claude/settings.json` or to global settings with the `--global` flag, and every tool call is stored in SQLite. Question search sends the contents of logic files to the Anthropic API and turns on with `ANTHROPIC_API_KEY`. Classification by words such as MUST and NEVER is heuristic and can miss files. The last commit dates from October 2026. Ideas such as a diff viewer and PDF export are still planned.
How to disable. To remove the hooks, delete the claude-watch entries from `.claude/settings.json` or `~/.claude/settings.json` and stop the dashboard. SQLite data goes away with the tool project folder.
FAQ
How do I run Claude Watch?
Build the project with `npm install` and `npm run build`, start the dashboard with `node dist/cli.js start` and in your project run `node /path/to/claude-watch/dist/cli.js init`. The dashboard opens at `http://localhost:3853`.
What does the bubble map show?
The project logic files: system prompts, voice guides, rules, configs and playbooks. A click opens the file.
How do I turn on question search?
Set `ANTHROPIC_API_KEY` and restart `node dist/cli.js start`. Plain language questions then run across the logic files.
Which commands exist?
`claude-watch start` launches the dashboard, `init` sets up hooks in a project, `init --global` in all projects, `snapshot <name>` saves state and `status` shows server stats.
Does it show agent actions in real time?
Yes, Claude Code hooks send every tool call to the server and the dashboard shows it over WebSocket at once.
Discussion
No questions or reviews yet.
Related
CLI toolsClaude Usage
claude-usage
A local dashboard from Claude Code logs: tokens, models, sessions and projects. For Pro and Max subscribers it shows the full picture instead of a gauge
A Claude Code sample: a lead agent, pools of coders, DevOps and reviewers, a spec and four test tiers
A GitHub Action that runs Claude Code in pull requests and issues: answers mentions, reviews and makes changes
A terminal dashboard for Claude Code and Codex: a tree of agents with live status, a tool-call feed and context headroom.
A local dashboard for Claude Code, Codex CLI, Gemini CLI, Copilot and OpenCode that shows tokens, cost, latency and errors
Claude Code Tools
claude-code-tools
A set of practical tools for Claude Code, Codex CLI and similar agents: utilities, skills, agents, hooks and plugins for terminal work
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| Claude Watch (this card) | no | 69 | |
| claude-dashboard | no | 587 | another Claude Code dashboard, while Claude Watch draws a map of logic files and records calls through hooks into SQLite |
| Claude Code History Viewer | no | 2,314 | shows past conversations, while Claude Watch shows live calls and a map of project rules |
README in short
The README shows the dashboard and explains that Claude Watch finds the logic files of a project and draws them as a map. A quick start, enabling question search, how hooks write calls to SQLite and broadcast them over WebSocket, a command table and contribution ideas follow.