We built OpenBrowser MCP to fix that.
Most browser MCPs give the LLM dozens of tools: click, scroll, type, extract, navigate. Each call dumps the entire page accessibility tree into the context window. One Wikipedia page? 124K+ tokens. Every. Single. Call.
OpenBrowser works differently. It exposes one tool. Your agent writes Python code, and OpenBrowser executes it in a persistent runtime with full browser access. The agent controls what comes back. No bloated page dumps. No wasted tokens. Just the data your agent actually asked for.
The result? We benchmarked it against Playwright MCP (Microsoft) and Chrome DevTools MCP (Google) across 6 real-world tasks:
- 3.2x fewer tokens than Playwright MCP - 6x fewer tokens than Chrome DevTools MCP - 144x smaller response payloads - 100% task success rate across all benchmarks
One tool. Full browser control. A fraction of the cost.
It works with any MCP-compatible client:
- Cursor - VS Code - Claude Code (marketplace plugin with MCP + Skills) - Codex and OpenCode (community plugins) - n8n, Cline, Roo Code, and more
Install the plugins here: https://github.com/billy-enrizky/openbrowser-ai/tree/main/pl...
It connects to any LLM provider: Claude, GPT 5.2, Gemini, DeepSeek, Groq, Ollama, and more. Fully open source under MIT license.
OpenBrowser MCP is the foundation for something bigger. We are building a cloud-hosted, general-purpose agentic platform where any AI agent can browse, interact with, and extract data from the web without managing infrastructure. The full platform is coming soon.
Join the waitlist at openbrowser.me to get free early access.
See the full benchmark methodology: https://docs.openbrowser.me/comparison See the benchmark code: https://github.com/billy-enrizky/openbrowser-ai/tree/main/be... Browse the source: https://github.com/billy-enrizky/openbrowser-ai
#OpenSource #AI #MCP #BrowserAutomation #AIAgents #DevTools #LLM #GeneralPurposeAI #AgenticAI
billy-enrizky-1•1h ago
We built OpenBrowser MCP to fix that.
Most browser MCPs give the LLM dozens of tools: click, scroll, type, extract, navigate. Each call dumps the entire page accessibility tree into the context window. One Wikipedia page? 124K+ tokens. Every. Single. Call.
OpenBrowser works differently. It exposes one tool. Your agent writes Python code, and OpenBrowser executes it in a persistent runtime with full browser access. The agent controls what comes back. No bloated page dumps. No wasted tokens. Just the data your agent actually asked for.
The result? We benchmarked it against Playwright MCP (Microsoft) and Chrome DevTools MCP (Google) across 6 real-world tasks:
- 3.2x fewer tokens than Playwright MCP
- 6x fewer tokens than Chrome DevTools MCP
- 144x smaller response payloads
- 100% task success rate across all benchmarks
One tool. Full browser control. A fraction of the cost.
It works with any MCP-compatible client:
- Cursor
- VS Code
- Claude Code (marketplace plugin with MCP + Skills)
- Codex and OpenCode (community plugins)
- n8n, Cline, Roo Code, and more
Install the plugins here: https://lnkd.in/e9GVAUDn
It connects to any LLM provider: Claude, GPT 5.2, Gemini, DeepSeek, Groq, Ollama, and more. Fully open source under MIT license.
OpenBrowser MCP is the foundation for something bigger. We are building a cloud-hosted, general-purpose agentic platform where any AI agent can browse, interact with, and extract data from the web without managing infrastructure. The full platform is coming soon.
Join the waitlist at openbrowser.me to get free early access.
See the full benchmark methodology: https://lnkd.in/ewMp65yG
See the benchmark code: https://lnkd.in/eHK5qZ-i
Browse the source: https://lnkd.in/e2dBG53W
See the Demo: https://lnkd.in/em7WfyYK
#OpenSource #AI #MCP #BrowserAutomation #AIAgents #DevTools #LLM #GeneralPurposeAI #AgenticAI