ArchestackGitHub ↗ (opens in a new tab)

Open-source alternatives to Perplexity

GPT Researcher leads with 81, ahead of Morphic (75) and Local Deep Research (63). Updated from public GitHub data.

The highest-scoring open-source projects in RAG, knowledge and search. They are picks, not exact replacements, so read the weaknesses before you switch. The same picks fit if you are leaving Perplexity AI.

Most stars
Vane 37,688 stars
Open-source alternatives to Perplexity
RankProjectScore
1GPT ResearcherResearch agent that writes cited reports from web and local documentsRAG, knowledge and search30.3k starsApache-2.081 out of 100
2MorphicAI answer engine with cited search results and generative UI componentsRAG, knowledge and search9.2k starsApache-2.075 out of 100
3Local Deep ResearchAgentic research tool that searches web, papers and your files, writes cited reportsRAG, knowledge and search9.4k starsMIT63 out of 100
#4VaneSelf-hosted answer engine with cited sources over SearXNGRAG, knowledge and search37.7k starsMIT57 out of 100
#5MAESTROMulti-agent research platform that writes long reports from documents and webRAG, knowledge and search1.5k starsAGPL-3.017 out of 100

Reviews

181 out of 100

GPT Researcher

Research agent that writes cited reports from web and local documents

30.3k starsApache-2.0last commit Oct 2026

GPT Researcher splits a query into research questions with a planner agent, scrapes sources in parallel, filters the context and writes a cited report. It works over web search (Tavily by default) and local files such as PDF, CSV, Excel, Markdown, PowerPoint and Word. It runs as a FastAPI server with a static or NextJS frontend, or as the gpt-researcher pip package.

Strengths

  • Covers both web research and local documents (PDF, CSV, Excel, Markdown, PowerPoint, Word)
  • Usable as a pip package, a FastAPI server with two frontends, or via Docker Compose
  • Supports MCP data sources, recursive deep research, and LangGraph or AG2 multi-agent workflows
  • Context filter falls back to local BM25 keyword ranking when no TypeSafe key is set

Weaknesses

  • Needs an LLM API key and a search API key (OpenAI and Tavily in the README examples)
  • Default Jev context filter requires a TypeSafe API key and a hosted call
  • Deep research is quoted at about 5 minutes and $0.4 per run with o3-mini
  • README calls it experimental and provided as-is, with no warranty
  • no GPU
  • Docker + Compose
  • Needs OpenAI API key, Tavily API key, Python 3.12+, TypeSafe API key (optional, Jev filter), Google API key (optional, image generation)
  • Models: OpenAI, OpenAI-compatible APIs via OPENAI_BASE_URL, Google Gemini (inline images)
  • port 8000
Documentation ↗ (opens in a new tab)Website ↗ (opens in a new tab)Repository on GitHub ↗ (opens in a new tab)
275 out of 100

Morphic

AI answer engine with cited search results and generative UI components

9.2k starsApache-2.0last commit Oct 2026

Morphic is a Next.js (TypeScript) AI search app that answers queries with cited sources and renders results as inline components streamed from a JSON spec. It supports OpenAI, Anthropic, Google, Ollama, Vercel AI Gateway and OpenAI-compatible models, with Tavily, SearXNG, Brave or Exa for search. Chat history lives in PostgreSQL, and Docker Compose bundles Postgres, Redis and SearXNG.

Strengths

  • Docker Compose bundles SearXNG, so no search API key is required
  • Many model providers, including local Ollama and OpenAI-compatible endpoints
  • Four search backends: Tavily, SearXNG, Brave, Exa
  • Built-in accounts via Better Auth or Supabase Auth, plus guest mode and shareable result URLs

Weaknesses

  • Compose stack needs PostgreSQL, Redis and SearXNG alongside the app
  • Needs at least one model provider API key or a local Ollama setup
  • Local development requires Bun; RAM and GPU requirements are not stated
  • Chat history, auth and file upload need extra configuration beyond the quick start
  • no GPU
  • Docker + Compose
  • Needs PostgreSQL, Redis, SearXNG, Tavily, Brave, Exa, Supabase
  • Models: OpenAI, Anthropic, Google, Ollama, Vercel AI Gateway
  • port 3000
Documentation ↗ (opens in a new tab)Repository on GitHub ↗ (opens in a new tab)
363 out of 100

Local Deep Research

Agentic research tool that searches web, papers and your files, writes cited reports

9.4k starsMITlast commit Oct 2026

Local Deep Research runs multi-step research on a question using an LLM (Ollama or any OpenAI-compatible endpoint) and search engines such as SearXNG, arXiv and PubMed, then writes a report with citations. Sources can be saved to a per-user SQLCipher-encrypted library that is indexed and searched in later runs. The web UI listens on port 5000 and offers several strategies, including a LangGraph agent mode, plus egress scopes that limit where a run may connect.

Strengths

  • Per-user AES-256 SQLCipher databases, including per-user LLM API keys
  • Egress scopes (Adaptive, Public only, Private only, Strict) restrict search and model traffic
  • README reports ~95% SimpleQA (n=500) with Qwen3.6-27B on one RTX 3090
  • Researched sources can be downloaded, indexed and searched alongside the live web

Weaknesses

  • Needs a separate LLM backend and SearXNG; pip install does not include them
  • Egress guard is a partial safeguard, not an air gap; relays to hosted models go undetected
  • Self-hosted SearXNG still forwards queries to upstream internet engines
  • Private SearXNG URLs are blocked by default since v1.10.3 without an operator opt-in
  • GPU optional
  • Docker + Compose
  • Needs Ollama or OpenAI-compatible LLM endpoint, SearXNG, Docker (optional)
  • Models: Ollama, OpenAI-compatible endpoints
  • port 5000
Repository on GitHub ↗ (opens in a new tab)
#457 out of 100

Vane

Self-hosted answer engine with cited sources over SearXNG

37.7k starsMITlast commit Sep 2026

Next.js answer engine (formerly Perplexica) that runs searches through a bundled SearXNG instance, then answers with citations using Ollama, OpenAI-compatible servers, OpenAI, Anthropic, Gemini or Groq models. Offers Speed, Balanced and Quality modes, web, discussion and academic sources, image and video search, file uploads and a search API. One container on port 3000; a slim image uses your own SearXNG.

Strengths

  • Single Docker image bundles SearXNG; no search API key needed
  • Local models via Ollama or any OpenAI-compatible server, plus cloud providers
  • Browser search-engine shortcut via /?q=%s and a REST search API
  • Slim image works with an existing SearXNG (JSON format and Wolfram Alpha enabled)

Weaknesses

  • No authentication yet; listed as an upcoming feature
  • Own-SearXNG setups must enable JSON output and Wolfram Alpha or searches fail
  • Tavily and Exa search backends are marked coming soon
  • Ollama on Linux must listen on 0.0.0.0 for the container to reach it
  • no GPU
  • Docker + Compose
  • Needs SearXNG (bundled in the default image), LLM provider (Ollama, OpenAI-compatible server or cloud API)
  • Models: Ollama, OpenAI, Anthropic Claude, Google Gemini, Groq
  • port 3000
Repository on GitHub ↗ (opens in a new tab)
#517 out of 100

MAESTRO

Multi-agent research platform that writes long reports from documents and web

1.5k starsAGPL-3.0last commit Apr 2026

Planning, Research, Reflection and Writing agents run research missions over uploaded PDF, Word and Markdown documents and web search, producing long reports with visible agent steps. Retrieval uses BGE-M3 embeddings in PostgreSQL with pgvector; any OpenAI-compatible API, including Azure OpenAI, can serve the models. Docker Compose stack on http://localhost with CPU and NVIDIA variants.

Strengths

  • Mission checkpoints allow pause, resume and writing-phase recovery
  • Local embeddings (BGE-M3) and pgvector; local LLMs via OpenAI-compatible API
  • Search providers: Tavily, LinkUp, Jina and SearXNG
  • CPU-only compose file plus automatic NVIDIA GPU detection

Weaknesses

  • 16 GB RAM minimum (32 GB recommended) and 30 GB disk
  • Alpha (v0.1.10-alpha); last commit April 2026
  • Dual-licensed AGPLv3 or commercial; proprietary use needs a paid license
  • First startup takes 5 to 10 minutes while models download
  • RAM ≥ 16 GB
  • GPU optional
  • Compose
  • Needs Docker Compose v2+, API key for an AI provider or an OpenAI-compatible endpoint, PostgreSQL with pgvector (in compose)
  • Models: OpenAI-compatible APIs, Azure OpenAI (GPT-5), BGE-M3 embeddings
  • port 80
Documentation ↗ (opens in a new tab)Repository on GitHub ↗ (opens in a new tab)