Skip to content

PinchTab

Go-based HTTP server that lets AI agents and scripts control Chrome through a local API.

Stars
10.5k
Forks
788
Licence
MIT
Language
Go
Last update
9 days ago
Type
Dev tool

What it is

PinchTab is a standalone HTTP server that gives AI agents direct control over Chrome. It runs as a small Go binary, typically as a background daemon on your own machine, so agent tools can share one browser control point. It can run headless or headed Chrome, manage multiple browser instances and profiles, and offers a dashboard.

  • Let an agent read a news page and pull out structured text without raw HTML or screenshots
  • Create a headed Chrome profile with saved logins and have an agent perform a task in it
  • Run site audits with screenshots, console errors and Core Web Vitals via the audit command
  • Compare two site versions visually with the compare command as a CI gate
  • Run browser instances inside Docker for sandboxed automation

Great for

  • Developers building AI agents that need to browse and extract page content
  • Local, single-user automation on a machine you manage
  • QA or testing teams running several isolated Chrome instances
  • Users on Raspberry Pi or other ARM64 machines, per the README's ARM64 support note

Not ideal for

  • Public-internet or multi-tenant browser services: the README says the dashboard, API and MCP server are privileged control surfaces not designed for untrusted users or direct public exposure
  • Users who want a turnkey hosted service: the README says it is not positioned as an internet-facing browser service
  • People who need a simple script-only library rather than a running server
Step-by-step guide: how to use PinchTabInstall, first run, examples, tips and FAQs

How to get it

  1. 1

    See the README

    You'll need: A locally installed Chrome or Chromium

    Install the control-plane server (macOS/Linux installer script)

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

    Install and start the daemon

    pinchtab daemon install
  2. 2

    Run it

    • Install the binary with the install script or from the latest release
    • Run pinchtab daemon install to start the server and a default headless Chrome instance
    • Point your agent or tool at http://localhost:9867
    • Optionally run bare pinchtab to see security posture and suggested next steps
  3. 3

    Get the most out of it

    • Use pinchtab audit for site-level checks including accessibility and Core Web Vitals
    • Use pinchtab compare with --fail-on-diff to gate releases in CI
    • Create named headed profiles with cookies or extensions for authenticated tasks
    • Run inside Docker for stronger isolation when automating locally
    • Review docs/guides/security.md before any non-local deployment

Downloads

Strengths

  • Small single binary, which the README describes as about 30MB and driving a locally installed Chrome or Chromium
  • Security defaults are local-first: loopback binding, sensitive endpoints disabled, and browsing limited to local sites until you opt in
  • Token-efficient text extraction, which the README says uses about 800 tokens per page

Trade-offs

  • Security setup is a real responsibility: exposing it beyond localhost means you must handle tokens, network boundaries and TLS yourself
  • By default it only browses locally hosted sites, so you must explicitly widen access for normal web use
  • Requires Go-based tooling knowledge for building or extending; README details beyond this are limited in the truncated text

How it compares

ProjectStarsTypeLevelUpdatedBest for
pinchtab (this)10.5kDev toolintermediate9 days agoDevelopers building AI agents that need to browse and extract page content
stagehand25.9kLibraryintermediate6 days agoDevelopers building browser agents that must work on sites they do not control
browser-harness18.8kDev toolbeginner4 days agoDevelopers who already use Claude Code, Codex or Cursor and want the agent to act in their own logged-in Chrome
browser-use120kLibraryintermediate4 days agoPython developers who want to prototype an AI agent that operates websites
CloakBrowser32.7kLibrarybeginner4 days agoPython or JavaScript developers already using Playwright who need automation to get past bot checks
Skill_Seekers15.3kDev toolintermediate1 months agoDevelopers who want an AI assistant to know a specific library's documentation

Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.