Practical overview
When it fits
Use Browser Harness when an LLM needs to work through a real browser session, especially for logged-in personal workflows or browser tasks that can benefit from reusable helpers. Use the MCP path when the controlling client supports MCP, and consider the documented cloud path when parallel browser execution is required.
Who it is for
- Developers building LLM-driven browser automation
- Teams or individuals using MCP-compatible coding agents
- People automating tasks in an existing logged-in browser session
- Builders creating reusable, domain-specific browser workflow helpers
Documented uses
- Connect an LLM agent to a real browser session
- Automate demonstrated web workflows such as file uploads and collecting video posts
- Create reusable helpers when a browser task needs a missing capability
- Expose browser-control functions to an MCP client
- Scale from local browser work to parallel cloud browsers
Requirements & limits
The repository is public and lists an MIT license. The documented local setup uses Python 3.12 and uv.
- The local workflow connects through an editable CDP websocket rather than describing a separate hosted browser as its starting point.
- The agent can add reusable helpers in its local workspace while the core harness remains protected.
- The stdio MCP server is documented for clients including Claude Code, Devin and Cursor.
- The repository describes Browser Use Cloud for parallel browsers with live previews, proxies, stealth and CAPTCHA solving.
- The evidence does not establish support for every browser, client configuration or workflow.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 2 October 2026
- Original website title
- browser-use/browser-harness · Official homepage title · retained in its source language

