Browser agent side panel for your own OpenAI-compatible LLM server (based on Claude in Chrome).
Users
10
store count
Rating
0.0 / 5
unchanged in 30 days
Ratings
0
total ratings
Store rank
#195K
#163K among extensions
Version
0.4.19
updated 9 Oct 2026
About Browser Renegade
Advanced fork of Claude in Chrome for all LLM.
Browser Renegade is an AI agent that lives in Chrome's side panel and works the web for you: it reads pages, clicks, types, fills forms, switches tabs and takes screenshots, across 17 browser tools. It is built on the Claude in Chrome agent and adds what the official one leaves out: your choice of model, no site blocklist, and a code sandbox on your machine.
YOUR MODEL, NOT OURS
• Connect any OpenAI-compatible model server, on your own computer or hosted online.
• Or download a built-in Qwen 3.5 model once (9B or 4B) and run it on your GPU through WebGPU, with no account and no server.
• Model and reasoning effort sit right in the chat box.
• No Claude account, no login, no Anthropic API.
NO BLOCKED DOMAINS
The official extension asks a remote classifier whether a site is allowed and refuses whole categories of sites, such as banking, news and forums. Browser Renegade has no classifier and no category blocklist: the agent works wherever you point it. Chrome enterprise policies still apply.
CODE RUNNER
The agent writes code and runs it in a sandbox inside the extension, offline, with nothing to install:
• Python with libraries for data analysis and charts, plus several other languages and SQL
• Creates spreadsheets, documents, presentations and charts in a workspace per chat
• Reads the PDFs, spreadsheets and Word files you attach, and can upload workspace files to a site's file field
MORE
• Modes for every model: Ask before acting, Act without asking, Plan only
• MCP connectors: add remote servers by URL, or sign in with OAuth
• Workspace viewer for tables, documents, PDFs, notebooks, audio and video
• Recent tasks saved on your device, with Continue to pick a chat back up
• Shortcuts, scheduled tasks and tab groups, as in the official agent
PRIVACY
Nothing goes to Anthropic: the official extension's API calls, account lookups, telemetry, remote-control bridge and uninstall survey are answered locally or blocked. Your prompts and pages go only to the model server you choose, or stay on your GPU with the built-in model. Anonymous usage statistics (feature names and counts, no page content, no prompts) go to chunks.md and can be turned off in Settings > Advanced.
GOOD TO KNOW
• Pick models with solid function calling and vision; the agent works from screenshots.
• Keep "Ask before acting" on for banking and other sensitive sites: web pages can contain prompt-injection attempts.
• Compiled languages (C, C++, Rust, Go, Java) are sent to godbolt.org or wandbox.org to compile. Sharing a file publishes it to a public host, and only when you ask.
• The built-in model needs Chrome 137+ with WebGPU: 9B needs about 7 GB of GPU memory, 4B about 4.5 GB.
Free, nothing gated. If it saves you time, you can support development on Ko-fi: https://ko-fi.com/vendoland
Unofficial and independent. Not affiliated with or endorsed by Anthropic.