Exa
6.7/10Exa is a web search and retrieval API built for AI applications and agents. It gives developers semantic search, page contents, deep research, and monitoring endpoints.
Data last updated November 2026 · Our methodology
Overview
Exa, which started out as Metaphor, runs its own web index and offers it through APIs built for large language model apps, RAG pipelines, and autonomous agents. Instead of just returning keyword links, it uses neural or semantic retrieval. Its endpoints cover search at several speed levels (instant, fast, auto, and deep), Contents for pulling the full text of pages, Answer for cited responses, Monitors for recurring searches, and an Agent API whose cost depends on effort level. Websets builds structured lists of companies or people, and Exa Connect and an MCP server let agents and coding tools reach Exa directly. Named customers include Cognition, HubSpot, monday.com, CodeRabbit, and Firefox. Pricing is pay-as-you-go with no subscription. Standard search costs $7 per 1,000 requests, lite and instant cost $4, and deep search costs $12 to $15. Extra results, page contents, and summaries are each charged at $1 per 1,000. The free tier gives $10 of credits that reset each month. An annual enterprise package costs $12,000 upfront in credits and adds 25 QPS and zero data retention. Custom contracts add SLAs, custom indexes, and up to 1,000 results per query. Exa says it holds SOC 2 Type II certification and offers a HIPAA BAA. It suits developer teams building agents and research tools. The main drawbacks are higher per-query list prices than budget search APIs, extra charges that add up quickly, a slow deep mode, and API churn: the /research endpoint was retired in 2026.
Our Rating
Who Exa is best for
Good fit
Exa is an API-first product with SDKs and an MCP server, built for developers working on AI apps.
SaaS companies such as HubSpot and monday.com use it to add web search to their own products.
Pay-as-you-go pricing with monthly free credits keeps the cost of getting started low.
Enterprise plans include SOC 2, zero data retention, SLAs, and custom indexes.
Pros & Cons
Pros
- Its semantic search matches queries by meaning, which works well for conceptual and research queries in RAG and agent pipelines.
- There is no subscription, and $10 in free credits arrives every month, so prototyping costs almost nothing.
- Exa states SOC 2 Type II certification, a HIPAA BAA, and zero data retention for enterprise customers.
- One API covers search, page contents, cited answers, monitoring, and an agent endpoint.
Cons
- List prices of $4 to $15 per 1,000 searches are higher than budget rivals such as Serper or Parallel Fast.
- Charges for page contents, summaries, and results beyond 10 make total costs harder to predict.
- Deep search can take several seconds, which is too slow for real-time chat.
- Endpoints such as /research have been retired and /findSimilar is deprecated, so existing integrations need migrating.
Key Features
Exa Pricing
Free
- No payment method required
- Roughly 2,500 lite searches
- $10 in credits that reset monthly
- One-time $10 bonus for completing onboarding
Enterprise (custom)
- SLAs and MSAs
- Custom indexes
- Volume discounts
- Custom rate limits
- Postpaid invoicing
- 1:1 onboarding and support
- Up to 1,000 results per search
Pay-as-you-go
- No subscription or minimum spend
- Prepaid credits billed per request
Frequently Asked Questions
Yes. Exa gives $10 in credits that reset each month, which covers about 2,500 lite searches, and no card is needed.
No. Standard usage is pay-as-you-go on prepaid credits. Enterprise customers can choose an annual $12,000 credit package or a custom contract.
Exa states it holds SOC 2 Type II certification. It also says it complies with GDPR and CCPA and offers a HIPAA BAA.
Zero data retention is available with enterprise plans, and under it queries and results are not stored or used for training.
Auto and fast searches cost $7 per 1,000 requests for up to 10 results. Each extra result is $1 per 1,000.
Yes. Exa offers an MCP server, Exa Connect, and Python and TypeScript SDKs.
Similar Tools
See all alternatives
Amazon's public cloud platform, with over 200 pay-as-you-go services for compute, storage, databases, networking and AI, used by startups through large enterprises and governments.
Anthropic's organization-wide Claude plan adds chat, Claude Code, Cowork agents, connectors, and enterprise governance such as SCIM, audit logs, and a compliance API.
Fin is Intercom's AI customer service agent. It resolves support conversations across chat, email, voice and social, either inside Intercom or on top of an existing helpdesk.