agent-os / shell
CLAUDE.md237 starsChanged 7 months agoApache-2.0View on GitHub
React-based web UI for browsing and managing AgentOS worlds. Backend requirement: The API server must be running on localhost:7777. The Vite dev server proxies /api/* requests to it. Claude Code can interact with the running app via Playwright MCP tools. This enables visual debugging and UI verification. Navigation: - browser_navigate - Go to a URL - browser_navigate_back - Go back - browser_click - Click an element by ref - browser_type - Type text into an input Inspection: - browser_snapshot -…
The file
We split this CLAUDE.md into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 51
Ensure the dev server is running: npm dev (port 7778)
- Distinctive: line 52
Ensure the API backend is running on port 7777
- Distinctive: line 53
Ask Claude to navigate to the app:
- Very specific: line 110
Use browser_snapshot over browser_take_screenshot when you need to interact with elements (snapshots include refs)
≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
AgentOS Shell
React-based web UI for browsing and managing AgentOS worlds.
Tech Stack
- React 19 with React Router 7
- TanStack Query for data fetching and caching
- Tailwind CSS v4 via
@tailwindcss/vite - shadcn/ui components (Radix primitives)
- Vite 7 with SWC for fast builds
- TypeScript 5.9
- Use
npm
Development
# Start dev server (runs on port 7778)
npm dev
# Regenerate API types from running backend
npm openapi:types
# Build for production
npm buildBackend requirement: The API server must be running on localhost:7777. The Vite dev server proxies /api/* requests to it.
Routes
| Path | Page | Description |
/ | Home | Landing with feature cards |
/explorer | Explorer Overview | World stats, def counts, recent defs |
/explorer/manifest | Manifest | Raw manifest viewer with tabs |
/explorer/defs | Definitions | Filterable defs table |
/explorer/defs/:kind/:name | Def Detail | Single definition view |
/explorer/plans/:name | Plan Diagram | Plan DAG visualization |
/workspaces | Workspaces Index | List of workspace trees |
/workspaces/:wsId | Workspace | Tree file browser |
/governance | Governance Index | Proposals list |
/governance/draft | Draft | New proposal editor |
Debugging with MCP Browser Tools
Claude Code can interact with the running app via Playwright MCP tools. This enables visual debugging and UI verification.
Starting a Debug Session
- Ensure the dev server is running:
npm dev(port 7778) - Ensure the API backend is running on port 7777
- Ask Claude to navigate to the app:
Navigate to http://localhost:7778
Available MCP Browser Commands
Navigation:
browser_navigate- Go to a URLbrowser_navigate_back- Go backbrowser_click- Click an element by refbrowser_type- Type text into an input
Inspection:
browser_snapshot- Get accessibility tree (best for understanding page structure)browser_take_screenshot- Capture visual screenshotbrowser_console_messages- View console logs/errorsbrowser_network_requests- See API calls made
Interaction:
browser_fill_form- Fill multiple form fieldsbrowser_select_option- Select dropdown optionsbrowser_press_key- Press keyboard keysbrowser_hover- Hover over elements
Debug Workflow Examples
Verify a page renders correctly:
Navigate to http://localhost:7778/explorer
Take a snapshot to see the page structureDebug an API issue:
Navigate to http://localhost:7778/explorer
Check network requests to see API calls
Check console messages for errorsTest user interactions:
Navigate to http://localhost:7778
Click on the Explorer card
Verify the URL changed to /explorer
Take a snapshot to confirm the page loadedInspect element state:
Take a snapshot
Look for the element ref in the output
Use browser_evaluate to inspect its propertiesTips
- Use
browser_snapshotoverbrowser_take_screenshotwhen you need to interact with elements (snapshots include refs) - Element refs (e.g.,
ref=e32) from snapshots can be used in subsequent click/type commands - Network requests show all API calls - useful for debugging data loading issues
- Console messages capture React errors and warnings
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

