[opensourcedex]

Scira is a free, open source ai interaction & interfaces project written in TypeScript and released under AGPL-3.0. It has 11,893 GitHub stars, 1,500 forks and 25 open issues, and was last pushed 1 months ago. On this registry it ranks #37 of 76 tracked projects in AI Interaction & Interfaces, with 5 head-to-head comparisons available. It gained 12 stars over the last 6 tracked days.

What is Scira?

What it is

Scira is a minimalistic AI-powered search engine and agentic research platform, formerly called MiniPerplx. It lives in the TypeScript and AI interaction ecosystem, and it is licensed under AGPL-3.0. The project helps users find information on the internet and cites the sources used to form the answer.

The concrete problem it addresses is slow or unsupported online research. A user can ask a question, upload a PDF, or paste a URL, and Scira can plan sub-tasks, retrieve live sources, cross-check evidence, and return a grounded answer with inline citations. This replaces a manual sequence of searching, opening links, comparing sources, and assembling notes.

Key capabilities

  • Agentic planning breaks complex questions into steps, selects models and tools, and executes multi-step workflows end to end.
  • Grounded retrieval returns answers with inline citations, so users can click a source and audit the evidence.
  • Search modes cover web, chat, X, stocks, code, academic, extreme, Reddit, GitHub, crypto, prediction, YouTube, Spotify, and Pro modes for connectors, memory, voice, and XQL.
  • Web search uses multi-query parallel retrieval with deduplication through Exa, Firecrawl, Parallel, and Tavily.
  • Extreme search uses an LLM-driven deep research agent with multi-step planning, code execution, and R2 artifact storage.
  • URL content retrieval extracts content from pages including tweets, YouTube, TikTok, and Instagram.
  • Lookouts schedule recurring research agents that monitor topics, track changes, and send email updates.

Who uses it and how

  • Researchers use Scira to ask a question and receive a cited answer instead of manually assembling evidence from search results.
  • Analysts use modes such as stocks, crypto, prediction, academic, and X to retrieve domain-specific context from live sources.
  • Developers and technical users use code, GitHub, and URL retrieval modes to inspect repositories, documentation, and web pages.
  • Teams or individuals can use Lookouts to monitor changing topics and receive scheduled email updates.
  • Self-hosters can bring their own models, connect custom tools, and tailor the workflow under the AGPL-3.0 license.

Getting started

Scira can be tried through the hosted site at scira.ai. The project can also be self-hosted, with configuration for AI model integration, web search, Redis, rate limiting, artifact storage, and email delivery as needed.

When to use it — and when not to

Scira is useful when cited, multi-source research matters and the user wants planning, retrieval, and verification in one interface. It is less suitable when a fully offline or single-service deployment is required, because it depends on external services such as Exa, Upstash, R2 storage, and model providers through Vercel AI SDK. The repository also lists several Pro modes and shows zero contributors, so some capabilities may be gated and maintenance may be limited.

project readme (upstream, from github) — read inline

Scira

Research at the speed of thought. The agentic research platform that plans, retrieves, and cites — so you can think faster.

Vercel OSS Program

Scira


Context.dev

Context.dev, One API to fetch the internet

Sign up for 500 credits free, no card required.


🔗 Try Scira at scira.ai

Ask DeepWiki

Powered By

Vercel AI SDK Exa AI Upstash
For AI model integration and streaming For web search and content retrieval For serverless Redis and rate limiting

How It Works

  1. Ask anything — Type a question, upload a PDF, or paste a URL. Pick a mode or let Scira decide for you.
  2. Scira plans & retrieves — The agent breaks your question into sub-tasks, searches live sources, and cross-checks the evidence.
  3. Get cited answers — Receive a grounded answer with inline citations. Click any source to verify it yourself.

Features

Core Capabilities

  • Agentic Planning — Breaks complex questions into steps, selects the right models and tools, then executes multi-step workflows end to end
  • Grounded Retrieval — Every answer comes with inline citations you can click to audit the evidence yourself
  • Extensible & Open — AGPL-3.0 licensed. Self-host, bring your own models, connect custom tools, and tailor everything to your workflow
  • Lookouts — Schedule recurring research agents that monitor topics, track changes, and email you updates

Search Modes (17 modes)

Mode Description
Web Search the entire web with AI-powered analysis
Chat Talk to the model directly, no search
X Real-time posts, trends, and conversations
Stocks Market data, charts, and financial analysis
Code Get context about languages and frameworks
Academic Research papers, citations, and scholarly sources
Extreme Deep research with multiple sources and analysis
Reddit Discussions, opinions, and community insights
GitHub Repositories, code, and developer discussions
Crypto Cryptocurrency research powered by CoinGecko
Prediction Prediction markets from Polymarket and Kalshi
YouTube Video summaries, transcripts, and analysis
Spotify Search songs, artists, and albums
Connectors Search Google Drive, Notion & OneDrive (Pro)
Memory Your personal memory companion (Pro)
Voice Conversational AI with real-time voice (Pro)
XQL Advanced X query language for tweet analysis (Pro)

Tools (28 tools)

Search & Retrieval
  • Web search — Multi-query parallel web search with deduplication using Exa, Firecrawl, Parallel, and Tavily
  • Extreme search — LLM-driven deep research agent with multi-step planning, code execution, and R2 artifact storage
  • Academic search — Search academic papers and research using Exa and Firecrawl
  • Reddit search — Search Reddit with configurable time ranges using Parallel
  • X search — Search X posts with date range filtering and handle inclusion/exclusion using xAI Grok
  • YouTube search — Search videos, channels, playlists with transcript extraction using Supadata
  • GitHub search — Search repositories with structured metadata extraction using Firecrawl
  • Spotify search — Search tracks, artists, albums, and playlists via Spotify Web API
  • URL content retrieval — Extract content from any URL including tweets, YouTube, TikTok, and Instagram
Financial & Market Data
  • Stock charts — Interactive stock charts with OHLC data, earnings, and news using Valyu, Tavily, and Exa
  • Currency converter — Forex and crypto conversion with real-time rates using Valyu
  • Crypto tools — Cryptocurrency data, contract lookups, and OHLC charts using CoinGecko
  • Prediction markets — Query Polymarket and Kalshi data with Cohere reranking using Valyu
Location & Travel
  • Weather — Current weather, 5-day forecast, air quality, and 16-day extended forecast using OpenWeatherMap and Open-Meteo
  • Maps & geocoding — Forward/reverse geocoding and nearby place discovery using Google Maps API
  • Flight tracking — Real-time flight status with departure/arrival details
Media & Entertainment
  • Movie/TV search — Search movies and TV shows with detailed cast, ratings, and metadata using TMDB
  • Trending movies — Today's trending movies from TMDB
  • Trending TV shows — Today's trending TV shows from TMDB
Productivity & Utilities
  • Code interpreter — Write and execute Python code in a sandboxed Daytona environment with chart generation
  • Code context — Get contextual information about programming topics using Exa Context API
  • Text translation — Translate text (and text within images) between languages using AI models
  • File query search — Semantic search over uploaded files (PDF, CSV, DOCX, Excel) with Cohere embeddings and reranking
  • Connectors search — Search connected Google Drive, Notion, and OneDrive using Supermemory
  • Memory tools — Save and search personal memories using Supermemory
  • Date & time — Current date/time in multiple formats with timezone support
  • Greeting — Personalized time-of-day-aware greetings

LLM Models Supported

  • xAI: Grok 3, Grok 3 Mini, Grok 4, Grok 4 Fast, Grok 4.1 Fast, Grok Code
  • OpenAI: GPT 4.1 (Nano/Mini/Standard), GPT 5 (Nano/Mini/Medium/Standard), GPT 5.1 (Instant/Thinking/Codex), GPT 5.2 (Instant/Thinking/Codex), o3, o4 mini, GPT OSS 20B/120B
  • Anthropic: Claude Haiku 4.5, Claude Sonnet 4.5, Claude 4.5 Opus, Claude 4.6 Opus
  • Google: Gemini 2.5 Flash (Lite/Standard), Gemini 2.5 Pro, Gemini 3 Flash, Gemini 3 Pro
  • Alibaba (Qwen): Qwen 3 (4B/32B/235B), Qwen 3 VL, Qwen 3 Max, Qwen 3 Coder (Small/Standard/Plus/Next), Qwen 3 Next 80B
  • Mistral: Ministral 3 (3B/8B/14B), Mistral Large 3, Mistral Medium, Magistral (Small/Medium), Devstral 2 (Small/Standard)
  • DeepSeek: DeepSeek v3, v3.1 Terminus, v3.2, R1, R1 0528
  • Zhipu (GLM): GLM 4.5, GLM 4.5 Air, GLM 4.6, GLM 4.6V, GLM 4.7, GLM 4.7 Flash
  • Cohere: Command A, Command A Thinking
  • MoonShot: Kimi K2, Kimi K2.5
  • Minimax: M1 80K, M2, M2.1, M2.1 Lightning
  • ByteDance: Seed 1.6, Seed 1.6 Flash, Seed 1.8
  • Arcee: Trinity Mini, Trinity Large
  • Others: Vercel v0 (1.0/1.5), Amazon Nova 2 Lite, Xiaomi Mimo V2 Flash, StepFun Step 3.5 Flash, Kwaipilot KAT-Coder-Pro V1

Built with

readme truncated — read the full docs on github

Frequently asked questions

Is Scira free to use?

Scira is open source under the AGPL-3.0 licence. There is no licence fee and no seat count — you can self-host it or, where the project offers one, pay a vendor for a managed version instead.

What does Scira do?

Minimalistic AI-powered search engine for instant answers

What is Scira written in?

Scira is primarily written in TypeScript. Its source is publicly available at https://github.com/zaidmukaddam/scira, and it has 11,893 GitHub stars.