What you can do
Hybrid retrieval
Vector similarity, graph traversal, and keyword search fused via Reciprocal Rank Fusion. Switch modes per query.
Namespaces
Isolated memory stores per tenant, project, or user. Reference them by ID or by your own external IDs.
Data integrations
Connect Google Drive, Snowflake, object stores, and more via OAuth. Ingestion runs continuously in the background.
Synthesized answers
Use
ask to get a generated answer with cited sources, or use recall to get raw matches for your own prompt.How you talk to it
| Surface | When to use |
|---|---|
TypeScript SDK (@deyta-ai/sdk) | Node, Bun, or browser apps. Typed, retries, cancellation. |
| REST API | Any language. Bearer-token authenticated. See the API Reference. |
| MCP endpoint | Connect Claude or another agent directly to a namespace via Model Context Protocol. |
| Console | Web UI for managing namespaces, API keys, integrations, identities, and observability. |
Who it’s for
App developers
You’re building a product on top of LLMs and need persistent, queryable memory without running databases.
Agent builders
You want your agent to remember past conversations and surface relevant context across sessions.
Data teams
You have content in Snowflake or object stores and want it retrievable by an LLM without an ETL project.
Platform engineers
You’re standardizing memory across many internal teams and want one tenancy boundary, observability, and access control.
Where to go next
Quickstart
Generate an API key and run your first remember/recall roundtrip in under five minutes.
Concepts
Namespaces, memories, ingestion, and retrieval — how the pieces fit together.
API Reference
REST endpoints for memory operations, namespaces, and integrations.
SDK
The TypeScript SDK — installation, configuration, and typed errors.