Browser Use

Browser automation tasks take 30-120+ seconds. The standard MCP stdio transport has timeout issues with long-running operations—connections drop mid-task.HTTP transport solves thisby running as a persistent daemon that handles requests reliably regardless of duration.

Install as a Claude Code plugin for automatic setup:

# Install the plugin /plugin install browser-use/mcp-browser-use

- Installs Playwright browsers on first run
- Starts the HTTP daemon when Claude Code starts
- Registers the MCP server with Claude

Set your API key(the browser agent needs an LLM to decide actions):

# Set API key (environment variable - recommended) export GEMINI_API_KEY=your-key-here # Or use config file mcp-server-browser-use config set -k llm.api_key -v your-key-here

That's it! Claude can now use browser automation tools.

For other MCP clients or standalone use:

# Clone and install git clone https://github.com/Saik0s/mcp-browser-use.git cd mcp-server-browser-use uv sync # Install browser uv run playwright install chromium # Start the server uv run mcp-server-browser-use server

Add to Claude Desktop(~/Library/Application Support/Claude/claude_desktop_config.json):

{ "mcpServers": { "browser-use": { "type": "streamable-http", "url": "http://localhost:8383/mcp" } } }

For MCP clients that don't support HTTP transport, usemcp-remoteas a proxy:

- Real-time task list with status and progress
- Task details with execution logs
- Server health status and uptime
- Running tasks monitoring

The web UI provides visibility into browser automation tasks without requiring CLI commands.

- Tasks Tab:Complete task history with filtering, real-time status updates, and detailed execution logs
- Skills Tab:Browse, inspect, and manage learned skills with usage statistics
- History Tab:Historical view of all completed tasks with filtering by status and time

- Run existing skills directly from the dashboard with custom parameters
- Start learning sessions to capture new skills
- Delete outdated or invalid skills
- Monitor running tasks with live progress updates
- View full task results and error details

The dashboard provides a comprehensive web interface for managing all aspects of browser automation without CLI commands.

Settings are stored in~/.config/mcp-server-browser-use/config.json.

mcp-server-browser-use config set -k llm.provider -v openai mcp-server-browser-use config set -k llm.model_name -v gpt-4o # Note: Set API keys via environment variables (e.g., ANTHROPIC_API_KEY) for better security # mcp-server-browser-use config set -k llm.api_key -v sk-... mcp-server-browser-use config set -k browser.headless -v false mcp-server-browser-use config set -k agent.max_steps -v 30

Environment Variables > Config File > Defaults

Environment variables use prefixMCP_+ section +_+ key (e.g.,MCP_LLM_PROVIDER).

Option 1: Persistent Profile (Recommended)

# Set user data directory mcp-server-browser-use config set -k browser.user_data_dir -v ~/.chrome-browser-use

Connect to an existing Chrome instance (useful for advanced debugging):

# Launch Chrome with debugging enabled google-chrome --remote-debugging-port=9222 # Configure CDP connection (localhost only for security) mcp-server-browser-use config set -k browser.cdp_url -v http://localhost:9222

mcp-server-browser-use server # Start as background daemon mcp-server-browser-use server -f # Start in foreground (for debugging) mcp-server-browser-use status # Check if running mcp-server-browser-use stop # Stop the daemon mcp-server-browser-use logs -f # Tail server logs

mcp-server-browser-use tools # List all available MCP tools mcp-server-browser-use call run_browser_agent task="Go to google.com" mcp-server-browser-use call run_deep_research topic="quantum computing"

mcp-server-browser-use config view # Show all settings mcp-server-browser-use config set -k <key> -v <value> mcp-server-browser-use config path # Show config file location

mcp-server-browser-use tasks # List recent tasks mcp-server-browser-use tasks --status running mcp-server-browser-use task <id> # Get task details mcp-server-browser-use task cancel <id> # Cancel a running task mcp-server-browser-use health # Server health + stats

mcp-server-browser-use call skill_list mcp-server-browser-use call skill_get name="my-skill" mcp-server-browser-use call skill_delete name="my-skill"

These tools are exposed via MCP for AI clients:

The main tool. Tell it what you want in plain English:

mcp-server-browser-use call run_browser_agent \ task="Find the price of iPhone 16 Pro on Apple's website"

The agent launches a browser, navigates to apple.com, finds the product, and returns the price.

Multi-step web research with automatic synthesis:

mcp-server-browser-use call run_deep_research \ topic="Latest developments in quantum computing" \ max_searches=5

The agent searches multiple sources, extracts key findings, and compiles a markdown report.

Deep research executes a 3-phase workflow:

┌─────────────────────────────────────────────────────────┐ │ Phase 1: PLANNING │ │ LLM generates 3-5 focused search queries from topic │ └─────────────────────────────┬───────────────────────────┘ ▼ ┌─────────────────────────────────────────────────────────┐ │ Phase 2: SEARCHING │ │ For each query: │ │ • Browser agent executes search │ │ • Extracts URL + summary from results │ │ • Stores findings │ └─────────────────────────────┬───────────────────────────┘ ▼ ┌─────────────────────────────────────────────────────────┐ │ Phase 3: SYNTHESIS │ │ LLM creates markdown report: │ │ 1. Executive Summary │ │ 2. Key Findings (by theme) │ │ 3. Analysis and Insights │ │ 4. Gaps and Limitations │ │ 5. Conclusion with Sources │ └─────────────────────────────────────────────────────────┘

Reports can be auto-saved by configuringresearch.save_directory.

All tool executions are tracked in SQLite for debugging and monitoring.

PENDING ──► RUNNING ──► COMPLETED │ ├──► FAILED └──► CANCELLED

During execution, tasks progress through granular stages:

INITIALIZING → PLANNING → NAVIGATING → EXTRACTING → SYNTHESIZING

┌──────────────┬───────────────────┬───────────┬──────────┬──────────┐ │ ID │ Tool │ Status │ Progress │ Duration │ ├──────────────┼───────────────────┼───────────┼──────────┼──────────┤ │ a1b2c3d4 │ run_browser_agent │ completed │ 15/15 │ 45s │ │ e5f6g7h8 │ run_deep_research │ running │ 3/7 │ 2m 15s │ └──────────────┴───────────────────┴───────────┴──────────┴──────────┘

Shows uptime, memory usage, and currently running tasks.

AI clients can query task status directly:

- health_check- Server status + list of running tasks
- task_list- Recent tasks with optional status filter
- task_get- Full details of a specific task

- Database:~/.config/mcp-server-browser-use/tasks.db
- Retention:Completed tasks auto-deleted after 7 days
- Format:SQLite with WAL mode for concurrency

Warning:This feature is experimental and under active development. Expect rough edges.

Skills are disabled by default.Enable them first:

mcp-server-browser-use config set -k skills.enabled -v true

Skills let you "teach" the agent a task once, then replay it50x fasterby reusing discovered API endpoints instead of full browser automation.

Browser automation is slow (60-120 seconds per task). But most websites have APIs behind their UI. If we can discover those APIs, we can call them directly.

Skills capture the API calls made during a browser session and replay them directly via CDP (Chrome DevTools Protocol).

Without Skills: Browser navigation → 60-120 seconds With Skills: Direct API call → 1-3 seconds

mcp-server-browser-use call run_browser_agent \ task="Find React packages on npmjs.com" \ learn=true \ save_skill_as="npm-search"

- Recording:CDP captures all network traffic during execution
- Analysis:LLM identifies the "money request"—the API call that returns the data
- Extraction:URL patterns, headers, and response parsing rules are saved
- Storage:Skill saved as YAML to~/.config/browser-skills/npm-search.yaml

mcp-server-browser-use call run_browser_agent \ skill_name="npm-search" \ skill_params='{"query": "vue"}'

Every skill supports two execution paths:

1. Direct Execution (Fast Path) ~2 seconds

If the skill captured an API endpoint (SkillRequest):

2. Hint-Based Execution (Fallback) ~60-120 seconds

If direct execution fails or no API was found:

Inject navigation hints into task prompt ↓ Agent uses hints as guidance ↓ Agent discovers and calls API ↓ Return data

Skills are stored as YAML in~/.config/browser-skills/: