Vane is a free, open-source AI answering engine — a self-hosted alternative to Perplexity that you run on your own infrastructure. Formerly known as Perplexica, the project has earned more than 37,000 stars on GitHub and remains actively developed, with its latest updates in 2026. Ask a question, and Vane searches the live web, reads the sources and writes a cited answer.
Vane uses SearXNG to query multiple search engines without tracking, then feeds the results to the language model of your choice — hosted models from OpenAI, Anthropic or Groq, or fully local models through Ollama for complete privacy. Focus modes tailor searches to academic papers, YouTube, Reddit, Wolfram Alpha and more, and a discovery feed surfaces interesting stories. Because everything is self-hosted, your questions and reading history never leave your network, which makes Vane a favorite for homelabs and privacy-focused teams. Conversations keep their context within a thread, so follow-up questions refine an answer instead of starting the search over. A built-in API mirrors the interface, so scripts and other apps can request researched answers programmatically.
Vane is completely free and open source under the MIT license. There is no paid tier; your only costs are the server you run it on and any hosted model API keys you optionally connect.
Vane is for self-hosters, researchers and privacy-minded users who love the AI-answer format but want it on their own terms — their own models, their own data, and no subscription.