LocalMind by Interra

Privacy-first local AI side panel — Ollama + MCP + semantic page search

Ollama Sidekick puts a privacy-first AI assistant in your browser's side panel. All inference runs locally through Ollama (https://ollama.com) on your own machine — no cloud APIs, no accounts, no subscriptions, no data collection. Ask questions about the page you're viewing. The side panel automatically picks up the page's text and uses it as context, so you can ask things like "summarize this," "what are the key arguments?", or "find the pricing." Highlight text first to focus the answer on a selection. Build a searchable local knowledge base. Mark pages as favorites and the extension crawls them on a schedule, indexing them into a local semantic store. The Search tab lets you find content across everything you've saved using natural-language queries — not keyword matching. Optional tool calling. If you run local Model Context Protocol (MCP) servers, configure them in the options page and the AI can discover and use their tools during chat. Built for users who can't or won't send page content to a remote model. Everything — chat history, page snapshots, embedding vectors — stays in your browser's local IndexedDB. No analytics, no telemetry, no remote calls. Requirements: • Ollama installed and running locally — https://ollama.com (free, open source) • Two models pulled once: a chat model like llama3.2 (or any model you prefer) and the embedding model nomic-embed-text • Ollama started with the extension's origin allowed — full setup steps shown in the side panel on first launch Open source. Published by Interra Development Group, LLC.