What is WebBrain?
WebBrain is a browser extension that connects a language model to the active tab. It reads page text, links, forms, tables and PDFs through an accessibility-tree representation. Ask mode handles read-only questions, Act mode performs browser actions, and Dev mode adds page inspection and debugging tools. Action runs can present a plan for approval before executing tools.
The extension is available through browser add-on stores or can be loaded from source. It supports a managed default model, cloud model APIs and local OpenAI-compatible servers. Saved workflows, scheduled tasks and page watches support repeated work. An optional MCP server lets external coding agents delegate tasks to the user's existing browser session, while retaining the extension's permission checks.
The extension is licensed GPL-3.0-or-later and is listed on the Chrome Web Store, Firefox Add-ons and Edge Add-ons. It runs on the model the user chooses: a local llama.cpp or Ollama server, a cloud API, or a managed default that needs no setup.
WebBrain screenshots
Explore the interface. Open a screenshot to view it at full size.
Source: official WebBrain project page.
Key capabilities
- Page analysis Reads page content and answers questions about text, links, forms, tables and PDFs.
- Browser actions Clicks, types, navigates, uploads and downloads, with per-site permission prompts before consequential actions.
- Visual interaction Uses screenshots and coordinate-based actions when the selected model supports images.
- Reusable workflows Saves successful runs for reuse, export and sharing, and supports scheduled tasks and conditional page watches.
- Developer inspection Exposes page source, styles, console and network diagnostics in Dev mode with supported model tiers.
- MCP delegation Accepts browser tasks from compatible AI clients and supports structured extraction using JSON Schema.
Where WebBrain fits
WebBrain fits page research, form-based tasks and recurring browser workflows performed in an existing signed-in session. Developers can also use its inspection mode or delegate browser work from a coding agent through MCP. Before deployment, the selected model's tool-calling support, image capabilities and context capacity need checking against the intended workload.
Browser choice affects functionality. Firefox lacks several Chrome debugging and interaction capabilities, and some application event handlers or client-side navigation may not work as expected. The Firefox MCP bridge is documented as not yet verified on a real installation. Deployment review should also cover site permissions, provider data flows and the optional research-sharing setting; the extension can maintain only one bridge connection at a time.
Install
git clone --filter=blob:none https://github.com/webbrain-one/webbrain.gitReviews of WebBrain
Be the first to share your experience with WebBrain.
Write a review
Loading review form…
Alternatives to WebBrain
agent-browseragent-browser is a native CLI with skills for AI agents to navigate pages, fill forms, extract data and take screenshots using accessibility-tree snapshots.
MCP Playwright ServerPlaywright MCP lets AI clients automate Chrome, Firefox, WebKit and Edge through accessibility snapshots instead of screenshots.
MCP Puppeteer ServerArchived MCP reference server that lets AI clients drive a Puppeteer browser: navigate, click, fill forms, take screenshots and run JavaScript.