logoLaunch Vault
  • Submit Project
  • Pricing
  • Browse Projects
  • Alternative To
  • Teardowns
Sign inSign up
Sign in
/
/
/
Editor’s picksREA (Reverse Engineer Anything)PhotoCraftArtCraft Crafting AppsUniversal ModderVane

All Categories

3D & Motion DesignAPIs & IntegrationsAR/VRArtificial IntelligenceBlockchain & CryptoBusiness AnalyticsCMS & No-CodeData Science & AnalyticsDatabasesDesign ToolsDeveloper ToolsDevOps & CloudE-commerceEducation TechFinance & FinTechGaming TechGraphics & IllustrationGreen TechHardwareHealth TechInternet of Things (IoT)Machine LearningMarketing ToolsMobile DevelopmentMusic & AudioNatural Language ProcessingOpen SourcePlatformsProductivityProject ManagementPrototypingRoboticsSaaSSales ToolsSecuritySEO & AnalyticsServerlessTesting & QAUI/UXVideo & Audio ToolsWearablesWeb DevelopmentWriting & Editing

Recommended Products

  • vectorize-image
  • BG Remove Pro
  • FluxKontext.Biz
  • JFIF to PNG
logoLaunch Vault

© 2026 Launch Vault. All rights reserved.

support@launchvault.dev
Follow us on Twitter
Powered by Open-Launch

Discover

  • Trending
  • Categories
  • Submit Project
  • Popular Collections
  • Agentic AI News

Resources

  • Pricing
  • Sponsors
  • Blog
  • Indie Tools
  • Links

Legal

  • Terms of Service
  • Privacy Policy

Free Tools

  • Traffic Checker
  • Domain Rank Checker
  • AI Traffic Checker
  • Open Graph Checker
  • PayPal Fee Calculator
  • Stripe Fee Calculator
  • Customer Acquisition Cost (CAC) Calculator
  • More tools →

Featured on

Launch Vault - Featured on Startup FameFirsto Top 1 Daily WinnerListed on Turbo0Featured on MagicBox.toolsFeatured on findly.toolsFeatured on toolfame.comFeatured on saasfame.comFazier badgeFeatured on TheSaaSDir

Friend links

SoPilotZ-Image.winIndie.Deals
  1. Home
  2. /Vane

Editor’s pick

Vane: Free, Open-Source Perplexity Alternative (Self-Host & Review)

A self-hosted AI answer engine with cited web search and local or cloud models.

Official repository
License
MIT
Platforms
Docker · Linux · macOS · Windows
GitHub stars
37,640 · snapshot 2026-10-23
Latest release
v1.12.2 · checked 2026-10-23

On this page

What is Vane?FeaturesSelf-hostLLMs & Ollamavs PerplexityBrowser searchAPIRelatedFAQ

By LaunchVault editorial · Official sources checked October 23, 2026

Vane (formerly Perplexica) is a free, open-source, privacy-focused AI answer engine that runs on your own hardware. It combines cited web search through SearXNG with your choice of local Ollama models or cloud model providers. Self-hosting the app does not make cloud model calls or web searches offline.

This documentation-based review helps you decide if Vane is right for you. We have not independently benchmarked or security-audited the application. The latest GitHub release checked on October 23, 2026 was v1.12.2; confirm the current version and image before installation.

What Is Vane? (Formerly Perplexica)

Vane is the current name of the Perplexica project: the former official GitHub repository redirects to ItzCrazyKns/Vane. The project is written primarily in TypeScript and released under MIT. It runs the answer interface on infrastructure you control, with SearXNG retrieving web results and a selected model producing an answer with cited sources.

The official repository showed 37,640 stars and 4,191 forks when checked on October 23, 2026. These are snapshots, not live counts or a measure of answer quality. The project is community maintained; inspect recent releases and issues before relying on a particular integration.

Sources: Vane README (reviewed revision) · Former Perplexica repository redirect · Vane MIT license · Vane GitHub releases · Live repository statistics

Key Features

Search modes trade latency for depth: Speed aims for quick answers, Balanced is the everyday option, and Quality is intended for deeper research. You can choose web, discussion or academic sources, narrow searches to a domain, and see cited results. Vane also offers image and video search alongside text results.

The model selector supports local Ollama and cloud providers such as OpenAI, Anthropic Claude, Google Gemini and Groq; you can mix providers for different tasks. SearXNG supplies the web-search layer. Contextual widgets cover weather, calculations and stocks. File uploads accept PDFs, text and images for question answering. Discover surfaces articles, while search history is stored locally by the instance.

Sources: Vane README (reviewed revision)

How to Self-Host Vane with Docker

The official bundled image includes Vane and SearXNG. Run the command below, then open http://localhost:3000 to choose models and configure any API keys. The named Docker volume persists Vane data. Docker is required; a free MIT license does not cover your hosting or optional cloud-model charges.

If you already operate SearXNG, use the official slim-latest image and set SEARXNG_API_URL to your instance. That SearXNG instance must allow JSON results and enable Wolfram Alpha. A third route is to clone the repository, run docker build -t vane ., and launch your locally built image using the README's volume and port flags. Check the repository before exposing the service beyond your own machine.

Scroll sideways to see all columns →

Vane self-hosting options documented by the official README
OptionWhen to use itSetup detail
Bundled imageFirst local installitzcrazykns1337/vane:latest includes SearXNG
Slim imageExisting SearXNG instanceUse itzcrazykns1337/vane:slim-latest with SEARXNG_API_URL; enable JSON and Wolfram Alpha
Source buildInspect or customize the imageClone ItzCrazyKns/Vane, build with docker build -t vane ., then run the image
Official bundled Docker image
docker run -d -p 3000:3000 -v vane-data:/home/vane/data --name vane itzcrazykns1337/vane:latest

Sources: Vane README (reviewed revision)

Which LLMs Does Vane Support?

Vane documents Ollama for local models and OpenAI, Claude, Gemini and Groq for hosted models. Choose the provider in setup; a cloud provider generally needs its own API key and may bill separately. With Ollama, the model can run locally, but a web-backed answer still needs network access for search results.

If Vane runs in Docker and cannot connect to Ollama on macOS or Windows, the README recommends http://host.docker.internal:11434. On Linux, use a reachable host address; the official troubleshooting instructions set OLLAMA_HOST=0.0.0.0:11434 in the Ollama service and restart it. Binding Ollama to all interfaces can expose it on your network, so restrict access before using that setting.

Sources: Vane README (reviewed revision)

Vane vs Perplexity

Scroll sideways to compare with Perplexity →

Vane and Perplexity: plans, workflow and usage conditions
CompareVanePerplexity
CostFree MIT app; hosting and cloud API costs may applyHosted free and paid plans
PrivacyInstance and local history under your control; external providers still receive data when selectedSearch and models run through a hosted service
ModelsOllama or supported cloud providersModels available through Perplexity's plans
SourcesCited results retrieved through SearXNGCited web answers
OperationYou deploy, update and troubleshoot itPerplexity operates the service

Perplexity is a hosted search product with free and paid plans. Vane gives you control over the app and model choice, at the cost of running and maintaining your own instance. Both can produce answers with sources; neither makes every answer correct. Perplexity is the easier choice if you want a managed service. Vane is worth considering when self-hosting or local models are requirements.

Compare all Perplexity alternatives

Sources: Vane README (reviewed revision) · Perplexity subscription plans

Use Vane as Your Browser Search Engine

Open your browser's search-engine settings, add a site search, and use http://localhost:3000/?q=%s as the search URL. Give it a name or shortcut, then try a query from the browser address bar. Replace localhost and the port if your Vane instance runs elsewhere; only use a remotely accessible endpoint you trust.

Sources: Vane README (reviewed revision)

Vane's API

Vane documents a search API for applications that want to submit queries and receive generated answers. Its official API documentation is the source for current endpoints, request fields and model selection. Check the documentation for the deployed version before writing a client; this review does not promise a stable API contract.

Sources: Vane search API documentation · Vane README (reviewed revision)

Related on LaunchVault

Compare self-hosted answer engines in the Perplexity alternatives guide below. The related LaunchVault project listings cover adjacent research and search workflows; they are not claims of feature parity with Vane. Follow each project's own documentation before choosing a replacement.

Sources: Vane README (reviewed revision)

Related research and search tools on LaunchVault

  • LumenAI

    A conversational library for researching documents; a different source workflow from web answer search.

  • Searchless.ai

    AI search visibility tooling for site owners, adjacent to search rather than a self-hosted answer engine.

FAQ

Common questions to help you decide if Vane is right for you.

You can host Vane and its search history yourself, and SearXNG mediates web search. If you select a cloud LLM, prompts are sent to that provider under its policies. Review hosting access and provider settings for your threat model.

Vane's code is available under the MIT license. Running a server and using paid cloud-model APIs may still cost money.

The official Perplexica GitHub repository now redirects to Vane. Vane is the project's current name, not a separate fork in this review.

A local Ollama model can run without a cloud model provider, but live web search through SearXNG still needs network access. Do not assume ordinary web-backed answers work fully offline.

Not for a local Ollama model. Hosted providers such as OpenAI, Claude, Gemini or Groq require their own credentials and may charge for usage.

For users willing to maintain a self-hosted answer engine, it can cover a similar cited-search workflow. Choose Perplexity when a managed service matters more than local control.

Vane uses SearXNG as its web-search layer. The standard Docker image bundles it; the slim image expects a separately configured SearXNG instance.