Memory for agents.Instant.
Write JSON. Get a graph, semantic search, and queryable schema — automatically.
No pipeline. No separate stores. No glue code.
Agent tooling changes.
Memory should not.
RushDB is a durable memory layer for AI agents: structured records, relationships, semantic search, and schema discovery from the JSON your application already produces.
Typical agent stack
Memory split across systems.
App records, session state, embeddings, metadata, tool output, and relationships drift across separate stores.
Application memory layer
Application DB
facts, history, metadata
Redis
session state, cache
Vector DB
embeddings, similarity search
Your code owns:
sync · embed · index · join · retry
With RushDB
Write JSON or CSV. Get Records.
Store messages, tool results, documents, entities, and events as Records with the structure your workflow needs.
No forced memory schema. Send nested JSON as often as your workflow produces it. RushDB keeps structure, typed fields, searchable text, and relationships together.
Change the stack without rebuilding memory.
Memory lives outside any one model, framework, MCP client, or provider session, so the next agent can continue from the same durable context.
Durable memory layer
Every provider reads the same records, relationships, and searchable context.
SDK, MCP, and application workflows share one durable memory layer.
Keep context available without replaying transcripts or migrating session state.
Running in minutes.
Install
Add the SDK for TypeScript, Python, or use the REST API directly — no extra dependencies.
Push
Write any JSON object. RushDB parses field types, builds the graph structure, and indexes vectors server-side on the same write.
Recall
Query by meaning, by graph relationship, or both in a single call using the same SearchQuery shape every time.
Choose the workflow you are building.
Persistent memory, graph-aware retrieval, and AI app data without stitching together a database, cache, vector store, and sync jobs.
The glue code RushDB removes.
RushDB turns messages, tool outputs, documents, entities, and events into queryable context without a separate graph pipeline, vector sync job, or agent-specific storage layer.