Vane

by Kushagra SrivastavaAppResearch something

Runs your own AI search engine that answers questions with links to its sources

About Vane

Vane, formerly Perplexica, is an AI answering engine you run yourself. It searches the web and writes an answer with links to its sources, works with a local or cloud model, and answers questions about your files.

For Researchers and studentsPeople who run home serversLocal AI users

Made by Kushagra Srivastava

An open-source project by developer Kushagra Srivastava, published on GitHub under the MIT license. The README says it is built on the idea that AI should be easy for everyone to use.

Stars: ★ 38k
License: MIT
Cost: Usage-based
Platforms: Mac · Win · Linux
Runs: Self-hosted

Screenshots

From the project's own repository · 2 images, swipe or use the arrows.

A Vane answer in the browser with source cards above it and related images and videos beside it — Image: Kushagra Srivastava (MIT). Original file

The Vane home screen with an Ask anything box, a speed mode menu and weather and news cards — Image: Kushagra Srivastava (MIT). Original file

Overview

Vane, formerly called Perplexica, is a self-hosted AI answering engine that searches the web and writes an answer with links to the sources it used. You start it with one Docker command, open it in your browser and connect a local model through Ollama or a cloud provider such as OpenAI, Anthropic, Google Gemini or Groq. It has speed, balanced and quality modes, can focus on discussions or academic papers, and answers questions about files you upload. It suits people who want a private alternative to hosted AI search.

The Docker image bundles the SearxNG search engine, so you can start without a search API key, and your search history stays on your own machine.

Try these first

Start with the first one: it is a small task you can check easily.

When did [public event] take place? Cite your sources.First try
Summarise recent academic papers on [topic], with links.academic sources
Using only [website domain], how do I [task]?
Read this PDF and list its three main findings: [upload file]sends file to model

Example prompts written by Agent Flows Hub.

Features

From the creator's documentation.

Feature	What it does	Runs
Cited answers	A written answer with numbered links to the pages it used	Web search
Search modes	Speed, balanced and quality modes for quick or deeper answers	Your server
Source focus	Search the web, discussions or academic papers, or limit to specific sites	Web search
File questions	Upload PDFs, text files or images and ask about them	Your model provider
Images and video	Shows related images and videos next to the answer	Web search
Search history	Every search is saved on your server so you can come back to it	Your server

Good to know

Ask a research question and get a written answer with numbered links to the pages it drew on, from a search engine you run yourself. Switch to quality mode or academic sources for deeper questions.

Limits

There is no sign-in yet; authentication is listed as an upcoming feature. Keep it on your own computer or private network unless you add your own protection.
When Vane runs in Docker and Ollama runs on the same computer, the Ollama address must be host.docker.internal or the host's IP, not localhost.

Notes by Agent Flows Hub (original project name: Vane (formerly Perplexica)), written from the creator's documentation, not from running it ourselves. Vane README · Vane installation documentation

Stuck setting it up?

Tell us what happened. We take a look first; if it needs paid work, you get an estimate before anything starts.

Get help

How to start

Install itPaste the install request into an AI agent you already use, or run the official command yourself or follow the official guide.macOS, Windows, Linuxdocker run -d -p 3000:3000 -v vane-data:/home/vane/data --name vane itzcrazykns1337/vane:latest
Open itFollow the first-run steps in the app.
Give it a first taskAsk a question you already know the answer to, such as when a public event took place, and open two of the cited sources to check the answer.

Steps follow the creator's instructions. The request tells your agent to read them first and wait for your OK.

Before you start

A computer with macOS, Windows or Linux.
Docker installed and running
A local model through Ollama, or an API key from a provider such as OpenAI or Anthropic
Pay per use. The AI service it calls charges for what you use.

Details

Made by: Kushagra Srivastava
License: MIT Terms
Platforms: macOS, Windows, Linux
Runs on: Self-hosted server

View source Official docs Report a problem

For AI agents

This page as a Markdown install brief:

/en/library/vane/install.md

Other apps

Vibe Research Gives you one local workbench to follow and research China, Hong Kong and US stocks
VoiceStudio Clones voices, dubs videos and reads text aloud on your own computer
Voyager Adds folders, a timeline and export to Gemini, ChatGPT, Claude and other AI chat sites
World Monitor Shows world news, conflicts, markets and infrastructure on one live map with AI briefs
yovoice Turns text into speech, cloned voices and songs on your own computer, with no API fees
AgentFlux Runs document workflows and AI agents on your own computer, with results saved locally

See all 113 apps

GitHub