Exa API

Neural search API that performs semantic understanding of queries against a real-time web index, returning full page content rather than snippets. Implements multiple latency profiles (instant <180ms, fast ~450ms, auto ~1s) by trading off result quality and synthesis depth, allowing developers to optimize for speed or comprehensiveness. Uses neural embeddings to match query intent rather than keyword matching, enabling AI agents to find contextually relevant content across millions of indexed pages.

Multi-step research capability that performs iterative web searches and synthesizes results into structured JSON outputs, optimized for complex queries requiring comprehensive analysis. Latency ranges from 2-60 seconds depending on research depth, with built-in support for extracting structured data (e.g., company information with CEO name, founding year) directly from web sources. Enables AI agents to decompose complex research tasks into multiple search iterations and consolidate findings into machine-readable formats without post-processing.

Offers Zero Data Retention (ZDR) option for privacy-sensitive applications, ensuring that queries and results are not logged or retained by Exa. Enables compliance with privacy regulations (GDPR, CCPA) and data protection requirements by preventing query data from being stored on Exa infrastructure. Available as an enterprise option with custom pricing, suitable for applications handling sensitive user data.

Offers enterprise-grade content moderation and filtering options tailored to specific organizational policies and compliance requirements. Enables filtering of search results based on custom criteria (e.g., excluding certain content types, domains, or topics) without modifying the underlying search algorithm. Available as enterprise feature with custom configuration, allowing organizations to enforce content policies across all search operations.

Provides $1,000 worth of free API credits for startups and educational institutions, reducing barrier to entry for early-stage companies and academic research. Enables startups to build and scale AI applications using Exa without upfront costs, and allows educational institutions to use Exa for research and teaching. Grant program is separate from free tier (1,000 requests/month) and provides significantly more usage capacity.

Implements OpenAI SDK-compatible interface and native support for OpenAI function calling, enabling Exa to be used as a drop-in replacement for OpenAI search tools. Automatically formats Exa search as OpenAI tool schema and handles function calling protocol. Also supports Anthropic tool calling for Claude integration.

Provides enterprise-grade security features including SSO (Single Sign-On) for authentication, Zero Data Retention (ZDR) for privacy-sensitive deployments, and SOC 2 Type II compliance certification. Enables enterprise customers to meet security and compliance requirements without custom integration or data handling agreements.

Provides interactive API dashboard at dashboard.exa.ai with guided onboarding that generates stack-specific integration code based on user's technology choices. Dashboard handles API key generation, SDK installation, and provides code examples for selected framework/language combination. Reduces setup time from hours to minutes.

Retrieves complete webpage text and structured content from URLs with configurable crawl policies (daily, weekly, on-demand) and caching. Supports multiple content types (general web, code documentation, financial data) with separate pricing per content type. Enables RAG systems to maintain fresh, full-text indexes of web content without managing crawlers, and allows AI agents to fetch complete page context for a given URL without relying on search.

Monitors specified web content at scheduled intervals and delivers updates via webhooks, enabling AI agents to react to web changes without polling. Runs searches on a schedule (frequency configurable but not documented) and pushes results to a specified webhook URL when changes are detected. Priced at $15 per 1,000 requests, making it suitable for long-running monitoring tasks that would be expensive with polling-based approaches.

Generates concise, web-grounded answers to queries with streaming response support and built-in citations, optimized for sub-1-second latency. Combines search with synthesis to produce answers that cite their sources, enabling AI agents to provide factual, verifiable responses without requiring separate search + LLM synthesis steps. Priced at $5 per 1,000 requests, making it cost-effective for high-volume answer generation.

Dedicated search indexes optimized for specific content verticals: people lookup and enrichment, company data (70M+ companies with structured extraction), code search (GitHub repos, Stack Overflow, documentation with sub-200ms latency), and news. Each vertical uses domain-specific indexing and extraction logic, enabling AI agents to find and extract information from specialized sources more accurately than general web search. Company search includes structured output support for extracting fields like CEO name, founding year, and financial metrics.

Generates AI-powered summaries of web pages and intelligently highlights relevant sections to reduce token overhead for LLM processing. Achieves ~90% token reduction on example pages (Boeing Wikipedia) by extracting key information and highlighting relevant passages instead of passing full page text. Priced at $1 per 1,000 pages, enabling RAG systems to reduce LLM input costs while maintaining semantic completeness.

Provides native function calling bindings for OpenAI, Anthropic, and other LLM providers, enabling AI agents to call Exa search as a tool without manual schema definition. Implements schema-based function registry that automatically exposes Exa endpoints (search, deep search, contents, answer) as callable tools with proper parameter validation and response formatting. Integrates with Anthropic Tool Calling, OpenAI Tool Calling, and AI SDK by Vercel, eliminating boilerplate for tool integration.

Provides an MCP (Model Context Protocol) server that enables Claude and other MCP-compatible AI systems to access Exa search as a native resource. Implements the MCP specification for tool exposure, allowing Claude to discover and call Exa endpoints without explicit tool definitions. Includes dedicated MCP documentation and server implementation, enabling seamless integration with Claude Desktop and other MCP clients.

Provides official Python and TypeScript SDKs with framework-specific integrations for LangChain, CrewAI, LlamaIndex, and other popular AI frameworks. SDKs handle authentication, request formatting, response parsing, and error handling, reducing boilerplate for developers. Framework integrations provide pre-built components (tools, retrievers, agents) that work out-of-the-box with Exa endpoints, enabling rapid integration without custom adapter code.