Hyperspell

Hyperspell is a hosted memory backend for AI applications that need to give each end-user persistent, cited context across conversations without engineering an entire RAG platform in-house. Developers hit a single API to ingest data — files, URLs, connector syncs from Google Drive, Notion, Gmail, Slack, GitHub — and Hyperspell handles chunking, embedding, storage, permissions, retrieval, and reranking. Because retrieval is scoped by user identity, an agent asking "what did I discuss with the customer last week?" only ever sees documents that user has permission to see. Hyperspell layers semantic search, keyword filters, temporal reasoning, and a lightweight knowledge graph so results stay accurate even as the corpus grows into millions of chunks. Its OAuth broker handles the auth dance for third-party sources, so end-users click "Connect Notion", and Hyperspell manages token refresh, incremental sync, and revocation on the app's behalf — saving weeks of connector engineering. SDKs are available in TypeScript and Python and integrate cleanly with LangChain, LlamaIndex, Vercel AI SDK, and OpenAI Agents SDK. Pricing includes a free developer tier, usage-based pro tiers, and enterprise contracts with SOC 2, VPC deployment, and BYO-vector-store options. Hyperspell exposes an MCP server so an agent host can attach a user's memory as a first-class tool.