BrowserOps

About

ANY TASK IN YOUR BROWSER. DONE — WITHOUT YOU.

BrowserOps is an MCP-powered Chrome extension that lets Claude, Codex,
Cursor, and any MCP-compatible LLM take real actions in your browser.
Research, fill forms, extract data, schedule meetings, file expenses —
anything you'd otherwise do by hand.

Describe the task once. The agent plans it, executes it across tabs
and sites, and returns when it's done.

WHAT IT DOES

- Drives any website end to end — clicks, types, scrolls, submits
- Operates across multiple tabs and domains in a single task
- Reads pages, extracts structured data, writes results to your docs
- Streams its plan and steps so you can watch or walk away

WORKS WITH YOUR LLM

BrowserOps speaks MCP, the open protocol for connecting tools to
language models. If your client supports MCP, it works:

- Claude (Desktop and Code)
- OpenAI Codex / ChatGPT
- Cursor
- Any custom MCP client

YOU STAY IN CONTROL

- Optional approval prompts for sensitive actions (banking, email,
  admin panels) — opt in via the extension popup
- Disconnect from the toolbar at any moment — the agent can't act
  until you reconnect
- Full action log for every step the agent takes
- Per-tab and global rate limits prevent runaway runs

USE IT FOR

Research, monitoring, data entry, form filling, job applications,
procurement, inbox triage, QA flows. Anything repetitive that lives
in a browser.

HOW IT WORKS

BrowserOps connects to a small Node bridge running locally on your
machine (127.0.0.1:57321, token-authenticated). Your MCP client
sends commands to the bridge, which forwards them to this extension
over a localhost WebSocket. No remote servers. No telemetry.

GET STARTED

1. Add to Chrome
2. Run "npm i -g browserops" then "browserops init"
3. Connect your MCP client (one-line config)
4. Send your first task