Repository navigation
WebBrain — open-source browser agent extension with WebMCP support #2399
esokullu
started this conversation in
Show and tell
Replies: 1 comment
|
For comparison with the extension approach, I maintain an Electron/Harness client, DeepDeck, that combines chrome-devtools-mcp with WebMCP. The browser has its own persistent profile, and the DevTools bridge is scoped to the agent's bound website tab. WebMCP discovery and invocation use a separate directory/call path. The DevTools tools are also used by a Builder: the agent explores a site, writes and verifies missing WebMCP tools, then reuses the saved tools in later tasks. Website registrations and generated tools retain distinct sources. Implementation walkthrough and screenshots. The profile difference matters here: this is a desktop browser environment, rather than access to the user's existing Chrome/Firefox extension session. |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Sharing in case it's useful to folks here: WebBrain is an open-source (MIT) browser agent that runs as a Chrome/Firefox extension, and we recently shipped WebMCP support (with real Chrome smoke tests in CI).
Different tradeoff from a CDP-driven setup: it runs inside the user's own browser session, so it works on already-logged-in pages without a separate profile or auth setup — which seems to be a recurring question here. Model-agnostic too (local via llama.cpp/Ollama, or any OpenAI-compatible API).
Repo: https://github.com/webbrain-one/webbrain — happy to answer anything about the WebMCP implementation side.
All reactions