pinchtab: Control Chrome for AI Agents

Overview

PinchTab provides a browser control plane for AI agents and developers. A local HTTP server and CLI expose Chrome navigation, page inspection, and interaction without requiring agents to work directly with raw HTML or screenshots.

It is suited to local automation that needs persistent browser state or multiple isolated instances. Remote and distributed setups are possible, but require operators to secure the service and its privileged control interfaces.

Key Features

• Control Chrome through a CLI or HTTP API, with an MCP server also available.
• Extract page text and inspect interactive elements for agent workflows.
• Run headless or headed browsers with persistent, isolated profiles.
• Manage multiple Chrome instances from a server-first control plane.
• Attach to externally managed Chrome instances in an advanced mode.
• Run browser audits and compare sites using screenshots and reported findings.
• Package the control software as a standalone Go binary; Chrome or Chromium is installed separately.

Use Cases

• AI-agent developers can automate navigation, form entry, and page reading through browser tools.
• QA and release teams can audit sites or compare staging and production pages.
• Developers can maintain separate logged-in browser profiles for different tasks.
• Operators can run parallel browsing tasks across isolated browser instances.

Getting Started

On macOS or Linux, install and start the local service with:

curl -fsSL https://pinchtab.com/install.sh | bash
pinchtab daemon install

Then point your agent or tool at http://localhost:9867. See the README and documentation for setup and API details.

Alternatives

• stagehand: Stagehand offers an SDK for model-driven browser tasks, while PinchTab exposes a local Chrome server and CLI for direct agent control.
• index: Index is an archived Python agent that reasons through web tasks, while PinchTab provides a Go HTTP API for agents to control Chrome.
• CloakBrowser: CloakBrowser focuses on fingerprint-modified Chromium for stealth automation, while PinchTab provides an HTTP interface to local Chrome.

Considerations

• PinchTab requires a locally installed Chrome or Chromium browser.
• The documented primary tested workflow is local macOS and Linux. Windows binaries are available, but Windows support is described as limited and best-effort.
• The dashboard, API, and integrations are privileged control surfaces, not intended for untrusted users or direct public-internet exposure. Non-local deployments require deliberate network and access controls.
• Browsing is restricted to locally hosted sites by default. Allowing broader access reduces the security protections of that default posture.