#2727th of 32 in RAG, knowledge and searchFor research, robotics and education

scira

AI search engine that plans multi-step research and answers with inline citations

Stars
11.9k
License
AGPL-3.0
Last commit
Aug 2026
Language
TypeScript
Stars, last 29 weeks+503
31 Mar 2026: 11,400 stars7 Apr 2026: 11,430 stars14 Apr 2026: 11,448 stars21 Apr 2026: 11,467 stars28 Apr 2026: 11,483 stars5 May 2026: 11,495 stars12 May 2026: 11,517 stars19 May 2026: 11,538 stars26 May 2026: 11,566 stars2 Jun 2026: 11,589 stars9 Jun 2026: 11,603 stars16 Jun 2026: 11,617 stars23 Jun 2026: 11,624 stars30 Jun 2026: 11,636 stars7 Jul 2026: 11,666 stars14 Jul 2026: 11,675 stars21 Jul 2026: 11,714 stars28 Jul 2026: 11,738 stars4 Aug 2026: 11,755 stars11 Aug 2026: 11,771 stars18 Aug 2026: 11,784 stars25 Aug 2026: 11,795 stars1 Sept 2026: 11,808 stars8 Sept 2026: 11,821 stars15 Sept 2026: 11,851 stars22 Sept 2026: 11,862 stars29 Sept 2026: 11,872 stars6 Oct 2026: 11,880 stars13 Oct 2026: 11,889 stars20 Oct 2026: 11,903 stars

11,400 on 31 Mar 202611,903 on 20 Oct 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

Scira is a Next.js app that takes a question, uploaded file or URL, breaks it into sub-tasks, searches live sources and returns an answer with clickable citations. It has 17 search modes (web, academic, Reddit, X, YouTube, stocks, crypto and others) and 28 tools, and it routes to hosted models from xAI, OpenAI, Anthropic, Google and others through the Vercel AI SDK. Self-hosting needs API keys for many third-party services.

Who it is for: Self-hosters wanting a cited, agentic AI search app over hosted models

Strengths

  • 17 search modes, including academic, Reddit, YouTube, stocks and prediction markets
  • Every answer carries inline citations linked to the retrieved sources
  • Many hosted model families via one SDK, including Claude, GPT, Gemini, Qwen and DeepSeek
  • Scheduled recurring research agents (Lookouts) email updates on tracked topics

Weaknesses

  • Depends on many third-party APIs (Exa, Firecrawl, Tavily, Valyu, Supadata, Daytona)
  • README lists no local model support such as Ollama or GGUF
  • Connectors, Memory, Voice and XQL modes are marked Pro
  • README is inconsistent on which API keys are required to run it

What it needs

  • no GPU
  • Docker + Compose
  • Needs Exa, Firecrawl, Tavily, Parallel, Valyu, Supadata, Daytona, Upstash Redis, Cloudflare R2, Supermemory, Cohere, ElevenLabs, Database (DATABASE_URL), Better Auth
  • Models: xAI Grok, OpenAI GPT, Anthropic Claude, Google Gemini, Qwen
  • port 3000

Also in RAG, knowledge and search

See all 32
Also in RAG, knowledge and search
RankProjectScore
1deeptutorSelf-hosted AI tutor with knowledge bases, guided learning, books and quizzes42k starsApache-2.082 out of 100
2LightRAGGraph-based RAG framework with a REST API server and web UI40.3k starsMIT77 out of 100
3RAGFlowRAG engine with deep document parsing, template chunking and agentic retrieval Live demo ↗ (opens in a new tab)92.6k starsApache-2.074 out of 100
#4MorphicAI answer engine with cited search results and generative UI components9.2k starsApache-2.074 out of 100
#5Open NotebookSelf-hosted NotebookLM alternative with podcasts and 20+ model providers40.4k starsMIT73 out of 100