Scira

What it does

Scira is an open-source, agentic AI search engine that breaks your question into sub-tasks, retrieves live web sources in parallel, and returns cited answers you can verify. It supports 17 search modes — from academic papers and Reddit threads to stock charts and prediction markets — and runs any of 50+ LLMs from OpenAI, Anthropic, Google, xAI, Mistral, and more.

Key features

• 17 search modes — Web, Academic, X, Stocks, Crypto, YouTube, Reddit, GitHub, Extreme deep-research, and more
• 28 built-in tools — web search, code interpreter (Daytona sandbox), file semantic search, maps, flights, TMDB, Spotify, and Supermemory connectors
• Full inline citations — every answer links directly to sources; click to audit
• Lookouts (scheduled agents) — recurring research jobs that email you updates on tracked topics
• Self-hostable — AGPL-3.0; bring your own API keys and deploy via Docker or Vercel one-click

Installation

Claude.ai web

Scira integrates via the Vercel AI SDK MCP package. Add the hosted instance at scira.ai or self-host.

Self-host (Docker)

git clone https://github.com/zaidmukaddam/scira
cp .env.example .env  # fill in API keys
docker compose up

Then open http://localhost:3000.

Self-host (Node.js / pnpm)

pnpm install
pnpm dev

Deploy to Vercel

Click the "Deploy with Vercel" button in the README and provide the required environment variables.

Supported hosts

Scira is a standalone web application; it uses @ai-sdk/mcp internally but is not distributed as an installable MCP server for desktop clients.