Send content into your Brain Graph and query what you've captured — programmatically. Every public endpoint below works with a single API key. To mint one, create a free account and open the API Keys page.
All requests authenticate with a Bearer API key. Upload endpoints need the documents:write scope; query endpoints need brain:read. Keys are created and revoked from the API Keys page in your account.
Upload a file (PDF, DOCX, TXT, MD, EML) as multipart form data. Returns a document id immediately; processing runs asynchronously in the background.
JSON alternative to the multipart upload — send plain text in the content field, or a base64-encoded file in pdfBase64. Useful for no-code platforms that can't post a file directly.
Send a public URL and the brain fetches the page, extracts the readable body, rehosts inline images, and ingests the article like any other document. Streams progress as Server-Sent Events.
Submit a YouTube video URL. The server pulls the transcript, resolves speakers, and ingests the video into the knowledge graph. Each video is processed at most once per user.
Send an audio file (mp3, m4a, wav, ogg, opus, webm, aac, flac, mp4) as multipart form data. The audio is transcribed with speaker diarization and language detection, then runs the standard entity-extraction pipeline. Up to ~150 MB per file. See the voice notes feature page for the conceptual overview.
Poll with the document id returned by an upload. Status moves through queued, processing, syncing, and analyzed. Once analyzed, the document is fully searchable and its extracted entities and facts are available.
List the separate brains in your account (Personal, Marketing, per client…), with their ids. Use a brain id with the brainId parameter on uploads or brainIds on search to target one brain. By default search spans all your brains. See the multi-brain feature page for the conceptual overview.
List the documents you have uploaded, newest first — filename, upload date, ingest status, source URL, and which brain each landed in. Pass query to filter by filename or notes, limit and offset to page. This is a lookup over your document library, not a search of the text inside those documents — use POST /api/brain/search for content.
Natural-language search across every fact, document, and relationship in your knowledge graph. Returns ranked results with source attribution. Pass brainIds to narrow to one or more named brains.
Look up a specific person, company, topic, or project by exact name. Returns the canonical entity with its uuid and metadata — useful for getting a stable id before calling another endpoint.
Machine-readable description of all Brain API endpoints, scopes, and async job-polling instructions — built for AI agents discovering the API at runtime.