Open-source Perplexity alternatives

GPT Researcher is an open-source autonomous agent that performs deep research on a given task and writes up a report with citations. It can work from web sources and from local documents, and it works with any LLM provider, aiming for factual, unbiased output rather than a quick chatbot answer.

The project positions itself against several limits of manual and LLM-only research: manual work can take weeks, models trained on old data may hallucinate, token limits make long reports hard, and a narrow source set leads to shallow or wrong conclusions. Drawing on Plan-and-Solve and RAG ideas, it splits research into planning and parallelized agent work to increase speed and keep behavior stable. Customization options let you tailor domain-specific research agents, and topics note MCP server support.

GPT Researcher is written in Python and Apache-2.0 licensed, with documentation in several languages and a website at gptr.dev. You run it yourself with your own model API keys, and it suits analysts, students and developers who want automated, source-backed reports on a topic.

Khoj is a personal AI app that you can run on your own machine or scale up to an enterprise cloud deployment. You can chat with local or online models, such as Llama, Qwen, Gemma, Mistral, GPT, Claude, Gemini and DeepSeek, and get answers drawn from the internet and from your own documents, including images, PDFs, Markdown, org-mode files, Word documents and Notion pages.

It can be reached from a browser, Obsidian, Emacs, a desktop app, a phone or WhatsApp. Users create agents with their own knowledge, persona, chat model and tools, set up automations that deliver personal newsletters and smart notifications by email, and use advanced semantic search to find relevant documents. It can also generate images and speak responses aloud. The README highlights benchmark results for retrieval and reasoning and introduces Pipali, a separate open-source AI coworker for your own computer.

Khoj is written in Python and licensed under AGPL-3.0. You can self-host it, with offline models via llama.cpp, or use the hosted app. It suits individuals and teams who want an AI assistant grounded in their notes and files while keeping a choice of model.

Vane is an AI answering engine that runs on your own hardware. It combines information from the internet with support for local language models through Ollama and cloud providers such as OpenAI, Anthropic Claude, Google Gemini and Groq, and returns answers with cited sources while keeping searches private. Repository topics also reference Perplexica.

Search modes trade speed for depth: Speed Mode for quick answers, Balanced Mode for everyday queries and Quality Mode for deeper research. Sources can be the web, discussions or academic papers, and web search is powered by SearxNG, which queries multiple search engines without exposing your identity. Image and video search, domain-restricted search and file uploads for questions about PDFs, text files and images are included.

Other features include widgets for quick lookups like weather, calculations and stock prices, search suggestions as you type, a Discover feed of articles, and a saved search history. Vane is written in TypeScript and released under the MIT license, and its architecture is documented for readers who want to understand how it works.

BrowserOS is an open-source browser designed to be used by AI agents. It is positioned as a secondary browser that sits alongside Chrome rather than replacing it, giving agents a place to carry out web tasks while you keep browsing normally. The project presents itself as an alternative to AI browsers such as ChatGPT Atlas, Perplexity Comet and Dia.

You can import logins, bookmarks and extensions from Chrome in one click so agents work with your real accounts, then connect agents such as Claude Code, Codex, Cursor or VS Code, which BrowserOS detects on your machine. Agents run in parallel in their own tabs. A live dashboard on the new tab page shows which site each agent is on, and every session is saved as a replayable video with a step-by-step action timeline.

Everything runs locally on your machine and the project is free under the AGPL-3.0 licence. Its topics point to local model tooling such as Ollama and LM Studio and it exposes tools through the Model Context Protocol. Typical tasks include posting to social media, clearing an inbox, updating a CRM or filing expenses. An enterprise offering is linked from the repository.

Morphic is an AI-powered search engine with a generative user interface. It answers questions with sources attached, and instead of returning plain markdown it renders richer inline components such as source-credited images, grids and headings built from a streamed specification.

It supports Quick and Adaptive search modes and lets you pick from several model providers, including OpenAI, Anthropic, Google and Ollama, as well as OpenAI-compatible endpoints. Search can run through Tavily, SearXNG, Brave or Exa. Chat history is stored in PostgreSQL, results can be shared through unique URLs, files can be uploaded, and Supabase Auth handles sign-in, with a guest mode for anonymous use.

Morphic is built with Next.js, React, TypeScript and Tailwind CSS under the Apache-2.0 licence. A Docker Compose setup starts PostgreSQL, Redis, SearXNG and the app together, so no separate search API key is required, although at least one AI provider key must be configured. It can also be deployed to Vercel.

Scira, formerly called MiniPerplx, is a minimalistic AI search tool that finds information online and cites where it came from. The project describes it as an agentic research platform that plans, retrieves and cites, and it is available as a hosted site at scira.ai.

You can start by asking a question, uploading a PDF or pasting a link. Behind it, the README credits the Vercel AI SDK for model integration and streaming, Exa AI for web search and content retrieval, and Upstash for serverless Redis and rate limiting. It is written in TypeScript as a web app you can run yourself with your own provider keys.

Scira is licensed under AGPL-3.0, so modified versions offered over a network must share their source. It suits developers who want a reference AI search app to run or adapt, and individuals who want a cited, minimalistic alternative to conventional search.

Browser Operator is an open-source AI browser that runs agents on the web on your behalf. It is a desktop application for macOS and Windows built to support research, analysis and automation, with processing done locally on your machine. The project describes itself as an open alternative to AI browsers such as ChatGPT Atlas, Perplexity Comet, Dia and the Microsoft Copilot Edge browser.

Its multi-agent platform uses specialized agents that work together on complex web tasks. You choose an AI provider in the settings, including OpenRouter, OpenAI, Groq or LiteLLM, and local models through Ollama allow fully offline operation. Through LiteLLM it is compatible with more than 100 models from OpenAI, Claude, Gemini, Llama and others. The repository topics also mention MCP client support and LangGraph.

Listed use cases include literature reviews and market research, shopping comparison and price monitoring, and business automation such as talent sourcing, lead generation and compliance audits. The system requirements list macOS 10.15 or Windows 10 (64-bit), 8 GB of RAM and 2 GB of free disk space. It is released under the BSD-3-Clause license.

Farfalle is an open-source AI-powered search engine, described by its creator as a Perplexity clone, that lets you search the web and get synthesized answers. It is aimed at developers and privacy-minded users who want to self-host this kind of search experience and choose their own model provider rather than depend on a closed commercial service.

It supports local models through Ollama, such as Llama 3, Mistral, Gemma and Phi-3, custom models through LiteLLM, and cloud models including Groq-hosted Llama 3 and OpenAI's GPT-4o. Search results can come from multiple providers, including Tavily, SearXNG, Serper or Bing, and the project supports chat history, an agent mode that plans and executes a search for better results, and expert search.

The stack uses Next.js for the frontend and FastAPI for the backend, with Redis for rate limiting and shadcn/ui components. Running it locally requires Docker and, for local models, Ollama; API keys for search and cloud model providers are optional depending on which features are used. Farfalle is written in TypeScript and released under the Apache-2.0 license, with a cloud-model-only live demo available.

Open Deep Research is an open-source alternative to the deep research features offered by OpenAI and Gemini. It is a web interface built with Next.js and shadcn/ui on top of the original Deep Research command-line project, turning that CLI tool into a visual app for starting and monitoring research tasks.

The system combines search via FireCrawl, web scraping and language models through OpenAI. It researches iteratively and recursively, uses the model to generate targeted search queries and follow-up questions, and runs multiple searches in parallel. Depth and breadth parameters control scope. Output is a detailed markdown report with findings and sources, viewable in a built-in markdown viewer, shown with real-time progress, and downloadable.

API keys for OpenAI and FireCrawl are required and are stored in HTTP-only cookies. You can try a hosted version at anotherwrapper.com with your own keys or host it yourself with Node.js. The project is MIT licensed and sponsored by AnotherWrapper.

Fyin is an open-source answer engine written in Rust, intended as an alternative to Perplexity AI that you can run on your own machine. Its stated motivation is a tool that runs locally, is open source, and delivers answers quickly. It is released under the AGPL-3.0 licence.

It can use a local model through Ollama or the OpenAI API, and it keeps a local vector database for fast search. Searching, scraping and answering run in parallel for speed, and websites are scraped locally. The number of search results to parse is configurable. For web search it needs a Bing API key, or the URL of a SearXNG or DuckDuckGo endpoint.

Fyin is run from the command line, either with cargo or in a Docker container built from the repository, with settings supplied through an environment file. The project lists a hosted version and a simple website as future ideas rather than existing features, so it is mainly for people who prefer to self-host their search assistant.