Scira: The Minimalist AI Search Engine for Grok, Claude, and Beyond

Scira does not crawl the web itself; it delegates. Think of it as an intelligent dispatcher for the internet. When you ask a question, Scira routes the request to the most suitable AI model—such as Grok, Claude, Gemini, or GPT—and leverages specialized search tools like Exa, Tavily, or xAI. It then aggregates the findings, synthesizes the information, and provides a clear response with full citations.

Originally developed under the name MiniPerplx, Scira focuses on removing the clutter typical of modern search engines. Built on the Vercel AI SDK, it offers a distraction-free interface that supports a wide array of LLMs while ensuring every answer is backed by verifiable sources.

Core Capabilities

• AI-Powered Search: Connect with Grok, Claude, Gemini, GPT, or various models hosted via Groq. Every query is processed by the "brain" best suited for the task.
• Web Search: Utilizing the Exa API, Scira performs multi-query searches with customizable depth to find the most relevant information.
• URL Extraction: Provide a direct link, and Scira will parse the page content to generate a concise summary.
• Reddit Search: The Tavily API allows Scira to pull specific discussions and community opinions from Reddit, filtered by timeframe.
• X (Twitter) Search: Integrated xAI real-time search finds posts based on specific dates or accounts.
• Deep Search: A multi-step reasoning process designed to tackle complex, multi-layered questions.
• Academic Search: Exa AI targets scholarly databases to surface research papers complete with abstracts.
• YouTube Search: Locate videos and retrieve metadata, captions, and specific timestamps.
• Movie and TV Lookup: The TMDB API provides detailed cast information, audience ratings, and show metadata.
• Stock Charts: Merges live financial data from yfinance with the latest market news via Tavily.
• Currency Conversion: Accesses real-time exchange rates through yfinance.
• Code Interpreter: A built-in Daytona sandbox executes Python code and generates visual charts.
• Weather Updates: Uses the OpenWeather API to provide current conditions and localized forecasts.
• Maps and Places: Integrates with the Google Maps API to find specific locations and nearby points of interest.
• Flight Tracking: Monitors real-time flight statuses using the Aviation Stack API.
• Translation: Leverages advanced AI models to provide high-fidelity text translation across multiple languages.
• Memory: Utilizes Mem0 AI to securely store and retrieve personal context for more personalized results.
• MCP Server Search: Enables discovery of Model Context Protocol (MCP) servers through the Smithery Registry.

Search Groups

To simplify the user experience, Scira categorizes its tools into specialized "Search Groups":

• Web: General-purpose internet searching.
• Memory: Recalls personal context (requires authentication).
• Analysis: Performs code execution and data processing.
• Chat: Direct, conversational interaction with AI models.
• X: Searches posts and trends from X.
• Reddit: Aggregates threads and user-generated opinions.
• Academic: Focuses on peer-reviewed research and papers.
• YouTube: Specifically targets video content.
• Extreme: Conducts deep, multi-source research for exhaustive investigations.

Supported Models

• xAI: Grok 3, Grok 3 Mini, Grok 2 Vision
• Google: Gemini 2.5 Flash (preview), Gemini 2.5 Pro (preview)
• Anthropic: Claude 4 Sonnet, Claude 4 Opus (with reasoning capabilities)
• OpenAI: GPT-4o, o4-mini, o3 (with reasoning capabilities)
• Groq: Qwen QwQ 32B, Qwen 3 32B, Meta’s Llama 4 Maverick

Technical Architecture

Scira is engineered for low latency and high reliability using a modern tech stack:

• Framework: Next.js with Tailwind CSS
• AI Integration: Vercel AI SDK
• UI Library: Shadcn/UI
• Search Engines: Exa.AI, Tavily
• Specialized Services: OpenWeather, Daytona, Google Maps, Aviation Stack, TMDB, Mem0
• Authentication & Database: Better Auth, Drizzle ORM

Self-Hosting and Local Setup

You can use Scira via the official site or integrate it into your workflow. For example, you can set it as your default Chrome search engine by adding a site search for https://scira.ai?q=%s with the shortcut sh.

Docker Deployment (Recommended)

1. Clone the repository and create a .env file from the provided .env.example. Enter your API keys.
2. Execute docker compose up. The application will be accessible at http://localhost:3000.

To run the image directly:

docker build -t scira.app .
docker run --env-file .env -p 3000:3000 scira.app

Node.js Installation

1. Obtain API credentials for OpenAI, Anthropic, and Exa (Exa is required for web search functionality).
2. Copy .env.example to .env.local and input your keys.
3. Run pnpm install followed by pnpm dev.
4. Navigate to http://localhost:3000 in your browser.

The Philosophy Behind Scira

Scira serves as a high-speed, lightweight interface between the user and the vast information on the internet. It eliminates advertising and algorithmic bloat, focusing exclusively on providing answers with "receipts." Whether you use the hosted version or deploy it on your own hardware, you maintain full control over your search experience.