Browserllama
Fork me

Use local AI in Chrome with Browserllama

Browserllama lets you ask questions about the current page using local models. Use Ollama or MLX for full local control, or switch to Chrome built-in AI on supported builds.

Start Setup Fix Common Issues

Screenshot

Current Browserllama popup UI.

Browserllama extension popup screenshot

Features

Focused, practical features for local-first browsing assistance.

Local-first prompts

Ask questions against page content without sending prompts to hosted SaaS APIs.

Provider choice

Switch between Ollama, MLX, and Chrome built-in AI directly from the popup.

Model selection

Pick your local Ollama or MLX model in popup settings and keep your preferred default.

Fast fallback docs

Use this page for setup and troubleshooting when local AI is not ready yet.

Setup

Choose one provider path (Ollama, MLX, or Chrome built-in AI).

Option A: Ollama

  1. Install Ollama from ollama.com.
  2. Start the server with the app or run ollama serve.
  3. Pull a model, for example ollama pull deepseek-r1:8b.
  4. Load extension source from src/ in chrome://extensions with Developer Mode enabled.
  5. Open Browserllama popup, keep provider on Ollama, and send a test prompt.

Option B: MLX

  1. Install MLX runtime: brew install pipx, pipx ensurepath, pipx install mlx-lm.
  2. Start server, for example: mlx_lm.server --model mlx-community/Qwen2.5-7B-Instruct-4bit.
  3. Default endpoint is http://localhost:8080/v1 (change popup endpoint only if you run a custom host/port).
  4. Open Browserllama popup and select provider MLX (OpenAI-compatible).

Option C: Chrome built-in AI

  1. Use a compatible Chrome build/channel and account setup.
  2. Enable current built-in AI flags in chrome://flags.
  3. Relaunch Chrome and select provider Chrome built-in AI in popup.
If your popup says provider is unreachable, jump to Troubleshooting.

Troubleshooting

Fast fixes for the setup failures users hit most often.

Ollama is not installed or not reachable

  1. Confirm Ollama is installed: run ollama --version in your terminal.
  2. Start server: run ollama serve and keep it running.
  3. Confirm default endpoint responds: open http://localhost:11434 in your browser.
  4. Check Browserllama provider is set to Ollama and model exists locally.
  5. If needed, configure CORS origins for your extension ID via Ollama environment variables.

Model is missing or replies fail

  1. List installed models: ollama list.
  2. Install one supported model such as deepseek-r1:8b or qwen2.5:7b-16k.
  3. Retry from popup after selecting the same model in the dropdown.

MLX provider is not reachable

  1. Confirm MLX is installed: run mlx_lm.generate --help.
  2. Start server: mlx_lm.server --model mlx-community/Qwen2.5-7B-Instruct-4bit.
  3. Verify endpoint: open http://localhost:8080/v1/models in your browser.
  4. In popup, select provider MLX (OpenAI-compatible) and confirm endpoint matches your server.
  5. If server runs on custom host/port, update popup MLX endpoint (for example http://localhost:9000/v1).

Chrome built-in AI provider is unavailable

  1. Use a Chrome build/channel that supports built-in AI APIs in your region and account setup.
  2. Go to chrome://flags and enable built-in AI related flags (names can change by version): Prompt API / Gemini Nano / optimization model flags.
  3. Relaunch Chrome after changing flags.
  4. In Browserllama popup, switch provider to Chrome built-in AI.
  5. If still unavailable, test in DevTools Console: typeof LanguageModel or window.ai?.languageModel.
Flag names and availability vary across Chrome versions. If one name is missing, search chrome://flags for "AI", "Prompt", and "Gemini Nano".

FAQ

Common Browserllama questions with direct answers.

What is Browserllama?

A Chrome extension that lets you ask AI questions about the active page using local providers.

Do I need Ollama to use it?

Not always. You can use Ollama, MLX, or switch to Chrome built-in AI if your Chrome supports it.

How do I install Ollama quickly?

Install from ollama.com, run ollama serve, then pull a model like ollama pull deepseek-r1:8b.

Which folder should I load in Chrome extensions?

Load the src/ folder from this repository in chrome://extensions.

Why do I get no response?

Check provider status in popup first, then verify Ollama/MLX server + model readiness or Chrome built-in AI availability.

What is the default MLX endpoint?

http://localhost:8080/v1. Browserllama probes availability with GET /models and uses POST /chat/completions.

Which Chrome experiments should I enable for local AI?

Enable current built-in AI related flags in chrome://flags, then relaunch and retest. Names vary by Chrome version, so search by keyword.