Browser and computer-use agents, ranked
Tools that let AI drive a browser or desktop: some are CLIs or MCP servers for your coding agent, others run web tasks on their own. They click, fill in forms and sign in on your behalf, so the trust column matters here too.
| # | Agent | FARS | Trust | Stars | License | Language | Last commit |
|---|---|---|---|---|---|---|---|
| 1 | Chrome DevTools MCP | 79 | 17/29 | 53.9k | Apache-2.0 | TypeScript | 2026-10-19 |
| 2 | Obscura Headless Browser | 75 | 19/29 | 29.1k | Apache-2.0 | Rust | 2026-10-19 |
| 3 | AIHawk – AI Browser Agent | 59 | 14/29 | 32.2k | MIT | TypeScript | 2026-10-15 |
| 4 | Agent Browser: Browser automation CLI for AI agents | 54 | 10/29 | 44.3k | Apache-2.0 | Rust | 2026-10-19 |
| 5 | Page Agent | 48 | 8/29 | 29.5k | MIT | TypeScript | 2026-09-06 |
| 6 | Browser Use | 47 | 7/29 | 118.5k | MIT | Python | 2026-10-17 |
| 7 | Stagehand: The AI Browser Automation Framework | 47 | 8/29 | 25.7k | MIT | TypeScript | 2026-10-19 |
| 8 | Nanobrowser Web Automation | 41 | 8/29 | 14.1k | Apache-2.0 | TypeScript | 2026-10-15 |
| 9 | TARS Multimodal Operator | 30 | 0/29 | 39.3k | Apache-2.0 | TypeScript | 2026-10-09 |
What qualifies: standalone, maintained tools of this kind that people search for by name. Plugins, skill packs and config layers built on top of them are left out and listed on the host tool's profile. The membership list is kept in our public repository, so every change is on record.
The order is the FARS total; stars only break ties, because popularity measures attention rather than whether something holds up. Two limits are worth knowing before you read the order. These are static reviews — source, documentation and repository metadata, with no runtime reproduction — which caps reliability, effectiveness and verifiability and puts the achievable maximum at 77 rather than 100. And a gap of a few points sits inside the noise of that method. Read the table as a shortlist worth investigating, not as a verdict; every score links to its full dimension breakdown and the evidence behind it. Review method →
1. Chrome DevTools MCP
Give coding agents direct Chrome automation, debugging, and performance-analysis capabilities.
Where it differs
Full mode exposes automation, debugging, network, performance, memory, and optional tool categories; --slim exposes only three tools for navigation, script execution, and screenshots, making it better suited to basic tasks and smaller contexts. The default deployment launches Chrome with a dedicated profile, while --autoConnect is intended for sharing manual browsing state on Chrome 144 or later and --browser-url or --wsEndpoint suits connections from a sandbox to an external browser. The MCP server targets coding-agent integration, whereas the accompanying CLI supports operation without MCP.
2. Obscura Headless Browser
A lightweight, stealth-capable browser engine for agent automation and web scraping without Chromium.
Where it differs
The project explicitly presents Obscura as a replacement for headless Chrome. Its README reports about 30 MB of memory, a 70 MB binary, an 85 ms page load, near-instant startup, and built-in anti-detection, compared with 200+ MB, 300+ MB, roughly 500 ms, about two seconds, and no built-in anti-detection for headless Chrome. Both are listed as compatible with Puppeteer and Playwright. Adoption should still be validated against representative sites because Obscura's independent renderer and Web API coverage can produce behavior different from Chromium.
3. AIHawk – AI Browser Agent
Open-source AI browser agent: describe tasks in plain English and it browses, clicks, types, and reads real websites to accomplish them.
Where it differs
The wiki section compares various AI browser agents, but the given README does not directly name a competitor. It discusses OpenAI Operator's availability in a wiki article, but we lack specifics here.
4. Agent Browser: Browser automation CLI for AI agents
A fast native Rust CLI for browser automation, designed for AI agents to navigate, interact, and extract data.
Where it differs
Compared to Playwright or Puppeteer, agent-browser is specifically designed for AI agents, offering a ref-based snapshot system, background daemon, and MCP integration, whereas Playwright/Puppeteer are more general-purpose libraries requiring more coding.
5. Page Agent
Embed a natural-language GUI agent that operates your web interface in place.
Where it differs
The README states that its DOM-processing components and prompts are derived from browser-use, and credits that project for its web-automation and DOM-interaction patterns.
6. Browser Use
Run AI-driven web tasks through a local or managed cloud browser.
Where it differs
The fully hosted Cloud path fits teams that want to submit tasks through an API and delegate both the agent and browser infrastructure. The CLI fits users who already work with Codex, Claude Code, Hermes, OpenClaw, Pi, or Cursor and primarily need browser access added to that agent. The Python library is the better fit when an application needs custom tools, structured output, model selection, or direct control of agent code; both the CLI and library can use either local or cloud browsers. For models, the project recommends the browser-optimized BU2 but also documents OpenAI, Anthropic, and Google routes, leaving adopters to balance task performance, latency, and budget.
7. Stagehand: The AI Browser Automation Framework
Control browsers with natural language and code for flexible, maintainable, reliable automation.
Where it differs
Compared to traditional automation tools like Selenium or Playwright, Stagehand lets you replace some low-level code with natural language, reducing maintenance overhead.
8. Nanobrowser Web Automation
A Chrome and Edge extension that runs multi-agent web automation with your own LLM API keys.
Where it differs
The project presents itself as a free alternative to OpenAI Operator. Its stated distinctions are use of your own API keys, per-agent model selection, and local-browser execution; the README does not provide feature benchmarks or a direct comparison table.
9. TARS Multimodal Operator
Use natural-language and vision-driven automation across computers, browsers, and MCP-connected tools.
Where it differs
The repository positions UI-TARS Desktop as a native desktop GUI agent and lists Midscene for browser use, but the supplied material provides no feature or performance comparison between them.
In scope, not ranked yet
These projects belong in this ranking, but their FARS review could not find enough public source evidence to score (for example, the repository ships no source, or its docs moved off-repo). A 0 would be a verdict FollowAgents never reached, so they are listed here instead of ranked.
- Cua: Cross-OS Computer Use and Virtualization Platform · ★ 30k — Scale computer-use 2.0 with open-source drivers, cross-OS fleets, and benchmarks.