schrute

Schrute

Universal Self-Learning Browser Agent — record browser interactions, learn API patterns, and replay them as optimized MCP tools.

What It Does

• Record — Open a browser, perform actions, and Schrute captures the underlying API calls
• Learn — The capture pipeline extracts parameters, detects auth patterns, and generates replayable skills
• Replay — Skills execute through a 4-tier optimization system, starting at browser-proxied and promoting to direct HTTP

Quick Start

# Run directly with npx
npx schrute serve

# Or install globally
npm install -g schrute
schrute serve

Installation

npm (recommended)

npm install -g schrute
schrute setup  # Installs Playwright Chromium

npx (no install)

npx schrute serve

From Source

git clone https://github.com/user/schrute.git
cd schrute
npm install
npx playwright install chromium
npm run build
node dist/index.js serve

Multi-Client Setup

Schrute works with any MCP client. Configure your preferred client:

Claude Code

Add to your project's .mcp.json:

{
  "mcpServers": {
    "schrute": {
      "command": "npx",
      "args": ["-y", "schrute", "serve"]
    }
  }
}

Or if installed as a plugin, use the built-in .mcp.json with ${CLAUDE_PLUGIN_ROOT}.

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "schrute": {
      "command": "npx",
      "args": ["-y", "schrute", "serve"]
    }
  }
}

Cursor

Add to .cursor/mcp.json in your project:

{
  "mcpServers": {
    "schrute": {
      "command": "npx",
      "args": ["-y", "schrute", "serve"]
    }
  }
}

Windsurf

Add to .codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "schrute": {
      "command": "npx",
      "args": ["-y", "schrute", "serve"]
    }
  }
}

Cline

Add to cline_mcp_settings.json:

{
  "mcpServers": {
    "schrute": {
      "command": "npx",
      "args": ["-y", "schrute", "serve"]
    }
  }
}

Generic (any MCP client)

Use stdio transport with the command: npx -y schrute serve

MCP Tools

Meta Tools

Tool	Description	Key Parameters
schrute_explore	Start a browser session on a URL	url (required)
schrute_record	Begin recording an action	name (required), inputs (optional)
schrute_stop	Stop recording and process captures	—
schrute_skills	List learned skills	site (optional filter)
schrute_sites	List known sites	—
schrute_status	Engine status and session info	—
schrute_dry_run	Preview a skill execution	skill_id (required), params (optional)
schrute_confirm	Approve a pending skill confirmation	token (required), approve (boolean)

Browser Tools (19 available)

browser_navigate, browser_click, browser_type, browser_snapshot, browser_fill_form, browser_press_key, browser_hover, browser_drag, browser_select_option, browser_file_upload, browser_handle_dialog, browser_tabs, browser_take_screenshot, browser_wait_for, browser_close, browser_resize, browser_console_messages, browser_network_requests, browser_navigate_back

Blocked for security: browser_evaluate, browser_run_code, browser_install

Dynamic Skill Tools

As you record actions, Schrute generates new MCP tools automatically. These appear with names like schrute_skill_<site>_<action> and can be called with their discovered parameters.

MCP Resources

URI	Description
schrute://status	Engine mode, uptime, active session info
schrute://skills	Skill catalog with redacted summaries
schrute://sites	Known sites with visit history

MCP Prompts

Name	Description	Arguments
explore-site	Guided workflow for exploring a website	url (required)
record-action	Guided workflow for recording a browser action	url (required), action_name (required)

Configuration

Environment Variables

Variable	Description	Default
SCHRUTE_DATA_DIR	Data directory path	~/.schrute
SCHRUTE_LOG_LEVEL	Log level	info
SCHRUTE_AUTH_TOKEN	Auth token for HTTP transport	—
SCHRUTE_NETWORK	Enable network mode (true/false)	false
SCHRUTE_HTTP_TRANSPORT	Enable HTTP transport (true/false)	false
SCHRUTE_HTTP_PORT	HTTP server port	3000

Config File

Settings persist in ~/.schrute/config.json. Manage with:

schrute config set server.authToken my-secret
schrute config get server

Precedence

CLI flags > Environment variables > Config file > Defaults

Security Model

Schrute enforces 9 security gates on every skill execution:

• Capability check — Is the tier's capability enabled?
• Domain allowlist — Is the target domain permitted?
• Method restriction — Is the HTTP method allowed?
• Path risk heuristics — Does the path look destructive?
• Rate limiting — Is the site being hammered?
• SSRF prevention — Is the target a private IP?
• Redirect validation — Does each redirect hop pass policy?
• Budget tracking — Are tool call limits respected?
• Audit logging — Is the execution properly recorded?

All newly-activated skills require one-time confirmation before first execution.

Claude Code Plugin

When used as a Claude Code plugin, additional features are available:

Commands

Command	Description
/schrute:explore <url>	Start browser exploration
/schrute:record	Record an action
/schrute:skills	List skills
/schrute:doctor	Run health diagnostics
/schrute:status	Check engine status

Agents

• skill-validator — Validates recorded skills for correctness
• exploration-guide — Interactive browser exploration assistant
• skill-debugger — Diagnoses failing skill executions

Skills

• schrute-usage — Comprehensive usage guide
• browser-tool-selection — Decision guide for choosing browser approach

Development

Prerequisites

• Node.js >= 22
• Playwright Chromium: npx playwright install chromium

Build & Test

npm run build          # Compile TypeScript
npx tsc --noEmit       # Type check only
npx vitest run         # Run tests
npm run dev            # Watch mode

Project Structure

src/
  core/        — Engine, config, policy, logging
  server/      — MCP stdio/HTTP, REST API, daemon
  browser/     — Playwright adapter, browser manager
  capture/     — HAR pipeline, auth detection, parameter discovery
  replay/      — Tier-based executor, dry run, audit log
  skill/       — Types, validator, skill lifecycle
  storage/     — SQLite database, repositories, migrations
  discovery/   — Cold-start API discovery

Before Publishing

• Update repository URL in package.json and .claude-plugin/plugin.json
• Verify npm pack --dry-run includes all necessary files
• Run full test suite: npx vitest run
• Run type check: npx tsc --noEmit

License

MIT