Open-Source Perplexity Alternative (2026, Free)

Identifying Perplexity · Mark source

Perplexity mark on a teal plate used to identify the commercial AI search product
Perplexity AI mark. Source · License: Perplexity AI trademark.

This hub lists open-source AI-powered answering engines that provide web search with cited sources, similar to Perplexity AI. Perplexity is named only to identify that product class. We copy licenses and GitHub star counts from official repositories. We do not rank search quality against Perplexity.

Quick take.

No verified options listed yet.

How we pick

Comparison

NameLicenseStarsSearch BackendLLM SupportKey Features

Listed projects: Vane (ItzCrazyKns/Vane), Farfalle (rashadphz/farfalle), and Scira (zaidmukaddam/scira). Other open AI search projects on this site currently live on ChatGPT alternatives.

Sourced snapshot

Field Vane (2026-10-17) Farfalle (2026-10-17) Scira (2026-10-17)
Repository github.com/ItzCrazyKns/Vane github.com/rashadphz/farfalle github.com/zaidmukaddam/scira
Stars / Forks 37,258 / 4,146 3,533 / 321 11,844 / 1,487
License MIT Apache-2.0 AGPL-3.0
Created 2024-04-09 2024-04-25 2024-08-07
Search backend SearxNG (Tavily/Exa coming soon) SearXNG, Tavily, Serper, Bing Multiple (web, academic, X, Reddit, GitHub, stocks, etc.)
LLM support Ollama (local), OpenAI, Claude, Groq Ollama (local), OpenAI, Groq, LiteLLM (any custom) OpenAI, Claude, Groq, and others (cloud)
Key features Privacy-focused, self-hosted, cited sources, chat history Perplexity clone, generative UI, Docker deployment, expert search 15+ search modes (web, chat, academic, code, stocks, crypto, Reddit, GitHub, YouTube, Spotify, etc.)
Homepage — farfalle.dev scira.ai
Quality vs Perplexity Not independently tested Not independently tested Not independently tested

Why Vane is on a Perplexity hub

“Open source Perplexity alternative” is a durable query. The most-starred open-source project we could source on 2026-10-17 from a public GitHub repository is Vane (ItzCrazyKns/Vane): MIT license, 37,258 stars, created 2024-04-09. It is a privacy-focused AI answering engine that runs entirely on your own hardware, combining web search (SearxNG) with local LLMs (Ollama) or cloud providers (OpenAI, Claude, Groq), and delivers answers with cited sources while keeping searches completely private. That is the same job family as a hosted AI search product, not a proof that it substitutes for Perplexity’s speed, UI, or data sources.

Perplexity is Perplexity AI’s hosted AI search product. We use the Perplexity mark only to identify the search. We did not copy Perplexity pricing or feature names into a quality table.

Farfalle: Apache-2.0 Perplexity clone

Farfalle (rashadphz/farfalle) is an Apache-2.0 open-source AI-powered search engine explicitly labeled “Perplexity Clone” in the README. 3,533 stars, created 2024-04-25. It supports:

  • Local LLMs through Ollama (llama3, gemma, mistral, phi3)
  • Custom LLMs through LiteLLM
  • Cloud models (Groq/Llama3, OpenAI/gpt-4o)
  • Multiple search providers: SearXNG, Tavily, Serper, Bing
  • Docker deployment setup
  • Chat history and expert search

Tech stack: Next.js frontend, FastAPI backend, shadcn/ui components, Redis rate limiting, Logfire logging. The repository includes a pre-built Docker image. Homepage: farfalle.dev.

Scira (formerly MiniPerplx, zaidmukaddam/scira) is an AGPL-3.0 minimalistic AI-powered search engine with 11,844 stars, created 2024-08-07. It offers 15+ specialized modes:

  • Web: Search the entire web with AI analysis
  • Chat: Direct model conversation, no search
  • Academic: Research papers, citations, scholarly sources
  • Code: Language and framework context
  • X: Real-time posts, trends, conversations
  • Stocks: Market data, charts, financial analysis
  • Reddit: Discussions, opinions, community insights
  • GitHub: Repositories, code, developer discussions
  • Crypto: Cryptocurrency research (CoinGecko)
  • YouTube: Video summaries, transcripts, analysis
  • Spotify: Songs, artists, albums search
  • Prediction: Markets from Polymarket and Kalshi
  • Extreme: Deep research with multiple sources
  • And more (Connectors, Memory, Voice, XQL are Pro features)

Powered by Vercel AI SDK. Homepage: scira.ai. The AGPL-3.0 license requires releasing modified source code if you provide the software as a network service.

Local versus hosted

The code is on GitHub. Vane, Farfalle, and Scira all support self-hosting with Docker or manual setup. If your requirement is “run an AI search engine on hardware I control,” start from the respective repository and follow the setup instructions. If your requirement is “a free web search box,” those repositories also link to hosted demo instances (Farfalle has farfalle.dev, Scira has scira.ai), which are different artifacts and are not what this hub ranks.

What we left out

Other open Perplexity-style projects (including Search4All, Perplexideez, Sensei, OpenPerPlex, and others) will join this list when we attach them explicitly. Vane is the most-starred MIT option. Farfalle is the Apache-2.0 Perplexity clone with multiple search backends. Scira is the AGPL-3.0 multi-mode search with 15+ specialized modes.

Self-reported benchmark data on each README is quoted on the respective project pages and labeled self-reported. They are not a Perplexity comparison.

How to choose

  • Need MIT-licensed privacy-focused self-hosted AI search with SearxNG: Vane (ItzCrazyKns/Vane).
  • Need Apache-2.0 Perplexity clone with multiple search providers (SearXNG, Tavily, Serper, Bing): Farfalle (rashadphz/farfalle).
  • Need AGPL-3.0 minimalistic AI search with 15+ specialized modes (academic, code, stocks, Reddit, GitHub, crypto, etc.): Scira (zaidmukaddam/scira).
  • Need a commercial-friendly permissive license for a SaaS product: Vane (MIT) or Farfalle (Apache-2.0); avoid Scira (AGPL-3.0 requires source disclosure for network services).

What “open” does not mean

Vane, Farfalle, and Scira code is open-source and not gated. That is a “take the code and deploy” story. It is still your responsibility to set up the search backend (SearxNG instance, Tavily API keys, Serper keys, etc.), LLM providers (Ollama local models or cloud API keys), and infrastructure (Docker, Redis, databases). “Free” on this hub means the code’s license is not a SaaS SKU; it does not mean zero DevOps effort. We have not published a one-click deployment recipe for any of them.

README performance data (search speed, answer quality, cost per query) stays vendor self-reported. Treating any of those as “beats Perplexity” would be editorial malpractice.

See also ChatGPT alternatives, closed LLM APIs, and AI Chat & Agents.

Listed options

FAQ

Is there a free open-source Perplexity alternative I can self-host?

Three sourced rows. Vane is MIT-licensed with 37,258 GitHub stars (2024-04-09 created): a privacy-focused AI answering engine with SearxNG search, local/cloud LLMs (Ollama, OpenAI, Claude, Groq), and cited sources. Farfalle is Apache-2.0 with 3,533 stars (2024-04-25 created): supports local (Ollama) and cloud models, multiple search providers (SearXNG, Tavily, Serper, Bing). Scira is AGPL-3.0 with 11,844 stars (2024-08-07 created): a minimalistic AI search with multiple modes (web, academic, code, X, stocks, GitHub, Reddit, etc.). "Self-host" means the code is downloadable; setup complexity varies.

Does it match Perplexity quality?

We have no sourced bake-off against Perplexity. Each project documents its own architecture and features. Search backend quality (SearxNG vs Tavily vs Serper) is not a ranking on this page.

What's the difference between Vane, Farfalle, and Scira?

License: Vane is MIT, Farfalle is Apache-2.0, Scira is AGPL-3.0. Stars: Vane 37K, Scira 12K, Farfalle 3.5K (2026-10-17 check). Features: Vane focuses on privacy and self-hosted AI with SearxNG; Farfalle is a Perplexity clone with multiple search providers; Scira offers 15+ specialized modes (web, academic, code, stocks, Reddit, GitHub, etc.). All support local/cloud LLMs. Not a quality ranking.

What is the license?

MIT on Vane (ItzCrazyKns/Vane) as of 2026-10-17. Apache-2.0 on Farfalle (rashadphz/farfalle) as of 2026-10-17. AGPL-3.0 on Scira (zaidmukaddam/scira) as of 2026-10-17. Not legal advice.

Can I use them commercially?

MIT (Vane) and Apache-2.0 (Farfalle) are permissive licenses that allow commercial use. AGPL-3.0 (Scira) requires releasing your modified source code if you provide the software as a network service. Not legal advice.

Do I need to run my own LLM?

No. All three support cloud LLM providers (OpenAI, Claude, Groq, etc.) via API keys. They also support local LLMs (Ollama, LiteLLM) if you prefer full privacy and control.

Sources & verification

Last verified 2026-10-17 · Last updated 2026-10-17· Edited by forkalt editors · Editorial policy

License notes are not legal advice. Read the original LICENSE. Trademark & disclaimer.