Ground coding agents in real code and current docs

Exa indexes documentation and API references, repositories and issue threads, changelogs and release notes, and technical writing across the open web, alongside a code index of working examples.

Exa Contents returns the passage that answers the query, without the surrounding page navigation, and Exa Highlights return only the relevant excerpt rather than the full documentation page. A Google provider snippet from the Auth.js docs, for example, is about 180 tokens when a full page is usually 2,400.

Yes. Exa Search has tiers from Exa Instant, under 250ms, through Fast at about 450ms and Auto at about one second, to Deep at 4-15 seconds for agentic research across more sources. Pick the tier per call, so lookups inside the loop stay fast and the hard question gets the depth.

Exa's coding use case is for builders and coding agents that need working code examples and current docs, issues, and release notes while writing code. Exa's inference-providers use case is serve-time grounding, where a provider runs Exa Search server-side so production models answer from today's web; Exa's model-training use case is post-training RL and eval loops run on Exa's index and its versioned snapshots. See Exa for inference providers and Exa for model training.

General search APIs rank pages for a person to read. Exa ranks for what an agent needs to answer, and returns the passage rather than the page, so the agent loads the working example or the changed line of documentation instead of the page around it.

Exa is SOC 2 Type II audited, and honors GDPR and CCPA data-subject rights. Zero data retention is available, meaning your queries and results are not stored or used for training. Single sign-on comes with an Enterprise plan. Terms are on the enterprise page.

Exa is priced per request, so an agent loop pays for the searches and content fetches it makes. Signing up is free, and current rates for each endpoint are on Exa's pricing page.