A privacy-first AI assistant that lives in your browser's side panel. All inference runs locally through Ollama — no data leaves your machine, no accounts, no subscriptions.
Chat — Ask questions about the page you're reading. LocalMind by Interra automatically picks up the current page's text and uses it as context, so you can ask things like "summarize this article", "what are the key arguments here?", or "find the pricing information on this page."
Search your saved pages — Mark pages as favorites and the extension crawls them on a schedule, building a local semantic index. The Search tab lets you find relevant content across everything you've saved using natural-language queries, not just keyword matching.
Knowledge base — Browse all your crawled pages in one place, with auto-generated summaries and the ability to drill into page sections.
MCP tool calling — If you run local MCP servers, LocalMind by Interra can discover their tools and let the AI use them during chat.
- macOS, Linux, or Windows with Chrome 114+
- Ollama installed and running locally
- Two Ollama models pulled (one-time setup):
- A chat model:
llama3.2,qwen2.5-coder,mistral-nemo, or any model you prefer - An embedding model:
nomic-embed-text(used for semantic search)
- A chat model:
# Pull a chat model (pick one you like)
ollama pull llama3.2
# Pull the embedding model (required for search)
ollama pull nomic-embed-textChrome extensions are a different origin than localhost. Ollama blocks cross-origin requests by default, so you must allow the extension's origin when starting Ollama.
During development (unpacked extension), use the wildcard — your local extension ID changes if the folder moves or on a different machine:
OLLAMA_ORIGINS="chrome-extension://*" ollama serveAfter installing from the Chrome Web Store, the extension ID is permanent and the same for every user. Lock it down to that specific ID for better security. The exact command is shown in the extension's Options page (right-click the toolbar icon → Options) with a one-click copy button.
If Ollama is already running, stop it first (
Ctrl+Corpkill ollama) then restart with the command above.
Option A — Chrome Web Store (coming soon)
Option B — Load manually (developer mode)
- Download or clone this repository
- Install dependencies and build:
npm install npm run build
- Open Chrome and go to
chrome://extensions - Turn on Developer mode (toggle in the top-right corner)
- Click Load unpacked and select the
build/chrome-mv3-prodfolder
Click the LocalMind by Interra icon in your Chrome toolbar. The side panel opens on the right side of your browser window and stays open as you browse.
Type a question and press Enter. The assistant automatically includes the current page's text as context. If you highlight text on the page before asking, that selection is used as the primary context.
- Model selector (top-right of the header) — switch between any chat models you have installed in Ollama. Your choice is remembered across reloads.
- Refresh button in the page context bar — re-reads the current page if you navigated somewhere new.
- Clear chat link (bottom of the input area) — starts a fresh conversation.
Click the Favorites bar below the page context area to expand it. Add the current page with the + Add current page button, or paste any URL.
Once you have favorites saved:
- Click Crawl all to immediately fetch and index all your saved pages.
- Crawling also runs automatically in the background once per day by default. You can change the schedule (or disable auto-crawl) in Settings.
- The extension follows links one level deep from each favorite (same domain only), so saving a site's homepage often picks up related pages automatically.
Pages that require you to be logged in won't crawl correctly — the background fetcher doesn't send your session cookies (by design, for privacy).
Type a natural-language query and press Enter (or click Search). Results are ranked by semantic similarity, not keyword matching — so "what does this company make?" will find relevant content even if your exact words don't appear on the page.
Results show a confidence score. Amber-bordered results with a "low confidence" badge are the best available matches but didn't cross the similarity threshold — treat them as suggestions.
Browse all crawled pages with their auto-generated summaries. Pages discovered by following links are shown nested under the favorite that led to them. Click any URL to open it in a new tab.
The status bar shows "CORS error" next to Ollama
Restart Ollama with OLLAMA_ORIGINS="chrome-extension://*" ollama serve. If you know your extension's specific ID (visible in chrome://extensions), use that instead of * for better security:
OLLAMA_ORIGINS="chrome-extension://YOUR_EXTENSION_ID" ollama serveThe status bar shows "Disconnected" Ollama isn't running. Start it with the CORS flag above.
Search returns no results after crawling
Open Chrome DevTools on the background service worker (chrome://extensions → Inspect views: service worker) and check the console for embedding errors. The most common cause is nomic-embed-text not being installed — run ollama pull nomic-embed-text.
Pages crawl but show "3 words, no content" The page probably requires JavaScript to render (single-page app). The background fetcher makes a plain HTTP request without running JS, so SPAs return empty shells. This is a known limitation.
The extension doesn't respond on chrome:// pages or the Chrome Web Store
Chrome blocks extensions from running on its own pages. This is expected — the extension will show an error message if you try to use it there.
My model disappeared from the dropdown If a model you had selected is removed from Ollama, the selector will briefly flash green and auto-switch to the next available model.
- All data (chat history, page snapshots, embeddings) is stored locally in your browser's IndexedDB.
- No data is sent to any remote server. Inference happens entirely on your machine via Ollama.
- The crawler only fetches URLs you explicitly mark as favorites.
- No analytics, telemetry, or tracking of any kind.
For the full policy, see the published privacy policy.
Published by Interra Development Group, LLC. Source code is open and inspectable in this repository.
"LocalMind by Interra" is an independent community extension and is not affiliated with, endorsed by, or officially connected to Ollama. The extension connects to your locally-running Ollama instance but is otherwise a separate project. Ollama does not appear to have published a formal naming policy for third-party integrations, and numerous community tools already use the name under similar terms. If you have questions about this, you can reach the Ollama team at hello@ollama.com.
Repository: github.com/Interra-Development-Group/localmind
For architecture, contribution guidelines, and product strategy, see ARCHITECTURE.md, CONTRIBUTING.md, and STRATEGY.md.
npm install
# Dev build with hot reload
OLLAMA_ORIGINS="chrome-extension://*" ollama serve &
npm run dev
# Load build/chrome-mv3-dev in chrome://extensions
# Production build
npm run build
# → build/chrome-mv3-prod/
# Package for Chrome Web Store submission
npm run package
# → build/chrome-mv3-prod.zipSee CLAUDE.md for architecture details, message passing contracts, and contribution guidelines.