Exa
Web search and retrieval API built for AI agents and apps
Overview
Exa (formerly Metaphor Systems) is a neural search and retrieval API built for AI agents and RAG pipelines, offering fast semantic web search, content extraction and monitoring; the SF-based company raised a $250M Series C at a $2.2B valuation in May 2026.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 11 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- 1,000 credits
- Up to 25 results/webset
- Limited features
- 8,000 credits/mo
- Up to 100 results/webset
- 2 team seats
- 100,000 credits/mo
- Up to 1,000 results/webset
- 10 team seats
- Custom credits
- Up to 5,000 results/webset
- Custom seats
What you can produce with Exa
- Build a RAG pipeline that retrieves semantically relevant web pages by meaning rather than keyword match, using Exa's neural search API to ground LLM outputs in real-time web content
- Construct an AI research agent that monitors specific web topics via Exa Monitors, receiving webhook alerts when new relevant content appears at a configurable cadence — for example, tracking competitor product announcements or regulatory updates
- Generate enriched B2B lead lists using Exa Websets: search for companies or people matching a semantic profile, then enrich records with contact emails and phone numbers via the built-in enrichment columns
- Extract clean, AI-optimised page text and highlights from any URL at scale using Exa's Contents endpoint, enabling automated ingestion of competitor content, news articles, or research papers into a data pipeline
- Power a deep research workflow using Exa Deep or Exa Deep Reasoning, which autonomously multi-hops across web sources to cluster and summarise findings about a topic before returning a structured, cited answer
- Find semantically similar documents or pages to a given URL using Exa's find-similar endpoint — useful for competitive analysis, content clustering, or surfacing related academic papers without knowing the right keywords
- Integrate real-time web grounding into an LLM chat application by wiring Exa's search-and-contents endpoints as a tool call, giving the model access to freshly crawled pages via Exa's continuously updated neural index
ASEAN Perspective
Exa in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
Exa earns its credibility through customer names, not just claims: Cursor, Notion AI, Cognition and OpenRouter all use it in production, and a $250M Series C at a $2.2B valuation (May 2026, led by a16z) backs that up. The API is clean, MCP-supported, and tuned for information-dense sources — research papers, code repos, company sites — which gives it precision edges over general crawlers on RAG and agentic-search tasks, though it competes hard with Tavily and Parallel on the same turf. Pricing is usage-based and can get complicated fast: search, contents, summaries and enrichments each bill separately, which developers flag as a legitimate cost-forecasting headache. There's no ASEAN-specific pricing or language tuning — it's English-first and globally available via API, useful for agent-grade retrieval rather than as a consumer search product.
What people say
A $2.2 billion valuation in May 2026, up from roughly $700M nine months earlier — Exa's Series C (a16z, $250M) tells you how fast the 'search engine for AIs' category has been moving. Formerly Metaphor Systems, rebranded in January 2024, Exa built its reputation on semantic rather than keyword search, and the customer list backs the pitch: Cursor, Notion AI, Cognition, HubSpot and OpenRouter are all named users in the company's own announcements, which counts for something even accounting for selection bias.
Developer feedback is consistently positive on integration speed and the quality of retrieval for nuanced, multi-hop RAG queries — this is where Exa's index, tuned for dense sources like papers, code and company sites, tends to beat general-purpose crawlers. Pricing sits at $7 per 1,000 search requests (raised from $5 in March 2026), with a free tier of 20,000 requests a month. The separate Websets product runs subscription tiers instead: free (1,000 credits), Core at $49/month (8,000 credits), Pro at $449/month (100,000 credits), and a custom Enterprise plan.
The most common developer gripe is billing complexity — search, contents, summaries and enrichment each meter separately, making cost forecasting harder than a flat per-query rate. No independently verified benchmark score against Tavily or Perplexity exists publicly as of mid-2026; comparisons vary by query type and evaluator, so treat any head-to-head claim, including Exa's own, with some caution.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including Exa's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Exa unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Exa directly →
Spotted something out of date? Suggest an update →
More in Research & Data
