AI Agents
macos-harness
A minimal harness that lets an LLM fully control a Mac.
★ 902 starsMIT
View guide →// agents · browser-use/browser-harness
browser-harness
browser-harness is an open-source ai agents project by browser-use (browser-use/browser-harness on GitHub). It has 18,355 GitHub stars and 1,794 forks, is released under the MIT license, and was last updated on GitHub on 27 September 2026.
A self-healing harness that connects an LLM directly to your real browser
★ 18.4k stars⑂ 1.8k forksMITupdated 2026-09
▦ visual — this project's README ships no embeddable screenshot,
so here's a placeholder. The guide below is complete regardless.
so here's a placeholder. The guide below is complete regardless.
// project facts
Project facts
- Repository
- browser-use/browser-harness
- GitHub stars
- 18,355
- Forks
- 1,794
- License
- MIT
- Last updated on GitHub
- 27 September 2026
- Category
- AI Agents
- Listing
- Full install guide
- Official site
- Official site
// how it works
How it works
- Decide local or cloud
Use your local browser for logged-in, personal work; Browser Use Cloud is for running many browsers in parallel.
- Get the harness
Paste the README's setup prompt into Claude Code or Codex and let the agent install it.
- Connect your browser
Enable remote debugging at chrome://inspect so the agent can attach over CDP.
- Run your first task
Give the agent a real browser task and watch it write any missing helpers on the fly.
// features
Why people love it
Direct CDP connection
Talks to your real browser through one editable CDP websocket - no separate automation browser profile required.
Self-healing helpers
When a capability is missing, the agent writes a reusable helper into its workspace, so repeat tasks get faster.
Skill-driven workflow
A SKILL.md teaches the agent the browser workflow; install.md handles connection setup.
MCP server included
browser-harness-mcp exposes the browser-control helpers as MCP tools over stdio for Claude Code, Devin, Cursor and other clients.
Works with your logins
Because it drives your real browser, tasks can use your existing logged-in sessions.
Cloud scaling path
Scale out with Browser Use Cloud for parallel browsers with live previews, proxies and stealth when local is not enough.
Advertisement
// run it
Run it yourself
- Clone the repository
Get the source code from GitHub and enter the project directory.
git clone https://github.com/browser-use/browser-harness.git cd browser-harness - Install via the setup prompt
The README's recommended route is pasting its setup prompt into your coding agent; it installs with uv on Python 3.12 and registers the skill.
# Paste the Setup prompt from the README into Claude Code or Codex - Connect the browser
The agent opens the remote-debugging settings page; tick the checkbox so it can connect.
# Open chrome://inspect/#remote-debugging and enable the connection as the README shows
// use cases
What you can build
- Download your latest 20 video posts from your X profile (the README's demo task)
- Automate a multi-step task on a site where you are already logged in
- Give Cursor or Claude Code real browser control through the MCP server
- Build a reusable helper library for your recurring web chores
Advertisement
// faq
Questions, answered
Who maintains this project?
It is maintained by the Browser Use team (browser-use) on GitHub.
Is it free to use?
Yes - Browser Harness is MIT licensed and free for local use. Browser Use Cloud is a separate paid service for scaling.
Does it need a paid API key?
You need an LLM agent (for example Claude Code or Codex) with its own model access; the local harness itself has no API fee.
How does it compare to alternatives?
Compared with Playwright scripts, the harness is agent-driven and self-healing - it writes missing helpers instead of failing - and it drives your real browser rather than a fresh automation profile.
How do I get started?
Paste the README's setup prompt into Claude Code or Codex, enable remote debugging when asked, and assign a small task.
// keep exploring
Similar tools
AI Agents
Jev Ultrafast
Jev Ultrafast — the fastest browser agent from browser-use: one decision-model request per action cycle drives a real browser, MIT
★ 23.7k starsMIT
View guide →AI Agents
reef
Reef — infrastructure for continually self-improving agents: weight-training and harness-evolving deployments.
★ 7.5k starsApache-2.0
View guide →AI Agents
webcmd
Self-learning browser infrastructure for AI agents — sites remembered, so agents stop rediscovering the web
★ 2.7k starsApache-2.0
View guide →AI Agents
open-kritt
Self-hosted AI vulnerability research: agents find, validate, deduplicate, and rank real security issues.
★ 2.2k starsAGPL-3.0
View guide →AI Agents
future-agi
One open platform to simulate, evaluate, guardrail, trace, and optimise LLM and agent apps, self-hostable.
★ 2.1k starsApache-2.0
View guide →// alternatives
Alternatives in AI Agents
The most-starred other AI Agents projects in the LibreAI catalogue — the realistic shortlist if browser-harness is not the right fit.
- openclaw★ 391.6k
A personal AI that really does things — on any OS, with your data staying yours.
- superpowers★ 294.9k
An agentic skills framework & software development methodology that works.
- ECC★ 275.2k
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
- hermes-agent★ 252.0k
The agent that grows with you — a Python framework for adaptive AI agents.
- deepseek-harness★ 245.5k
DeepSeek's plugin-based agent harness where everything is a plugin.
// more in agents
More in AI Agents
Nearby AI Agents projects in the catalogue, closest in stars to browser-harness — keep exploring the shelf either side of it.
- agent-lightning★ 18.6k
Microsoft's trainer for AI agents - reinforcement learning for almost any agent framework
- parlant★ 18.3k
Build customer-facing AI agents that actually follow your rules
- deepeval★ 18.7k
Unit tests for LLMs — evaluate before you ship.
- langchainjs★ 18.2k
The JavaScript/TypeScript version of LangChain - the agent engineering platform
- agent-zero★ 19.4k
Agent Zero AI framework
- openfang★ 18.2k
An open-source agent operating system in Rust - autonomous 'Hands' that do real work
Get browser-harness
Free and open — grab the source, star the repo, and start creating.