# PinchTab

> A standalone HTTP server and CLI that gives AI agents direct, token-efficient control over Chrome for browser automation and multi-instance orchestration.

PinchTab is an open-source, local-first browser automation tool built as a small Go binary (~16 MB) that exposes an HTTP API and CLI for AI agents to control Chrome. It supports both headless and headed Chrome profiles, persistent sessions, and multi-instance orchestration with strong isolation between parallel browser processes. Designed for token efficiency, PinchTab extracts page content at ~800 tokens per page — 5–13x cheaper than screenshot-based approaches — making it well-suited for agentic workflows where API cost matters.

- **Token-efficient extraction** — *Use `pinchtab text` to extract page content at ~800 tokens instead of 10,000, dramatically reducing LLM API costs.*
- **Headless and headed Chrome** — *Run Chrome without a visible window for automation, or with a visible window for tasks requiring authenticated profiles.*
- **Persistent profiles** — *Create named browser profiles with saved cookies, history, and local storage so agents stay logged in across restarts.*
- **Multi-instance orchestration** — *Spin up multiple isolated Chrome processes in parallel via the HTTP API, each with its own profile.*
- **HTTP API and CLI** — *Control the browser via `curl` or the `pinchtab` CLI; the server runs as a local daemon on port 9867.*
- **MCP server support** — *Expose browser tools to AI agents via the Model Context Protocol for seamless agent integration.*
- **Self-contained binary** — *Single ~16 MB Go binary with no external runtime dependencies; install via shell script, Homebrew, npm, or Docker.*
- **ARM64 support** — *First-class Raspberry Pi support with automatic Chromium detection.*
- **Local-first security** — *Binds to 127.0.0.1 by default, no telemetry, no analytics, and no required outbound service dependency.*
- **Dashboard** — *Real-time web dashboard for monitoring and managing instances, profiles, and tabs.*

To get started, install the binary with `curl -fsSL https://pinchtab.com/install.sh | bash`, then run `pinchtab daemon install` to start the background control plane. Point your agent or tool at `http://localhost:9867` and use the CLI or HTTP API to navigate pages, extract content, and interact with the DOM.

## Features
- Standalone HTTP API server for browser control
- Token-efficient page text extraction (~800 tokens/page)
- Headless and headed Chrome support
- Persistent browser profiles with saved sessions
- Multi-instance parallel Chrome orchestration
- CLI with interactive menu and shell completions
- MCP server integration for AI agents
- Real-time dashboard for monitoring instances and tabs
- Docker container support with volume persistence
- Local-first security with loopback binding by default
- ARM64 / Raspberry Pi support
- External Chrome attach via remote debug mode
- Accessibility-first stable element references
- No telemetry or analytics

## Integrations
Chrome, Chromium, Docker, Homebrew, npm, Model Context Protocol (MCP), Anthropic Claude, curl

## Platforms
WINDOWS, MACOS, LINUX, WEB, API, CLI

## Pricing
Open Source

## Version
v0.10.0

## Links
- Website: https://github.com/pinchtab/pinchtab
- Documentation: https://pinchtab.com/docs
- Repository: https://github.com/pinchtab/pinchtab
- EveryDev.ai: https://www.everydev.ai/tools/pinchtab