Memory for agents.Instant.
Write JSON. Get a graph, semantic search, and queryable schema — automatically.
No pipeline. No separate stores. No schema planning.
Push. Structure. Query.
Push data
Any JSON or CSV. Any shape. RushDB parses field names, types, and nested structure on write — nothing to define upfront.
Instant structure
Property names and types become graph nodes. Records sharing a property link automatically — no relationship declarations needed. Vectors indexed server-side on write, or bring your own embeddings.
Query everything
One unified SearchQuery for records, relationships, and schema. Combine semantic similarity and graph traversal in a single call — same shape every time.
Before and after.
Write Records. Query everything they imply.
One primitive. Relationships, vectors, and queryable schema emerge on write.
EPISODIC MEMORY
Facts and events, stored
Every tool output, conversation turn, and action — written once as a typed Record with automatic graph relationships between them. Retrieve by agent ID, time range, or semantic similarity.
Records + Relationships
SEMANTIC MEMORY
Meaning, auto-embedded
Any string property becomes a searchable vector — embedded server-side on write, no extra step. Query by meaning, not just by value. Bring your own model if you prefer.
Vector Properties + AI Indexes
STRUCTURAL MEMORY
Structure, instantly queryable
A live ontology of every label, property, value range, and relationship — built from your data, fetched in one call. Agents never construct queries blind.
Ontology API
THE TYPICAL STACK
Systems get stitched together
- — Redis for session state
- — Vector DB for search
- — Database for structure
Writes require syncing across systems.
Pipelines require maintenance
- — Embedding
- — Indexing
- — Schema updates
Every change forces a full reprocess.
Data drifts out of sync
- — Structure is partial
- — Relationships are missing
- — Indexes lag behind source data
Queries rely on assumptions instead of actual state.
RushDB makes structure, meaning, and relationships first-class — not outcomes of stitching systems together.
Who it’s for.
Every agent action, indexed and retrievable.
Store and retrieve agent state, tool outputs, and conversation history across sessions. Query by agent ID, time range, or semantic similarity — all in one call.
Why RushDB: No separate vector store. Graph links sessions, actions, and context automatically.
What's in the box.
Managed Embeddings
Mark any string property for embedding. RushDB embeds it server-side on every write — no embedding pipeline to run, no backfill jobs to schedule. Switch to your own model at any time by passing pre-computed vectors.
Vector + Graph
Semantic similarity and graph traversal compose in one query. Filter by relationships, rank by meaning — no separate stores to sync.
MCP Server
Add RushDB to any agent with one config block. Your agent can read, write, and semantically search your data mid-conversation — no extra setup, no custom integration.
ACID transactions
Two agents writing simultaneously won't corrupt each other's memory. Every write either fully completes or fully rolls back — no partial states, no silent data loss.
Self-host or Cloud
Run on your own infrastructure with Docker and your Neo4j instance, or use managed cloud with a free tier. No shared infrastructure, no lock-in.
Unified Query API
The same SearchQuery works for records, relationships, labels, properties, and deletes — identically. Learn it once on records. Never learn a separate paradigm for schema discovery, relationship traversal, or cleanup.
Works where you work.
Direct SDKs for every stack. Native MCP for every agent. Your coding agent learns RushDB from your dependencies, not your docs.
MCP SERVER
One config block. Every MCP agent.
Paste it once. Your agent can read, write, and query RushDB mid-conversation — records, relationships, semantic search, all of it.
WORKS WITH ANY MCP AGENT
AGENT SKILLS
Install expertise, not boilerplate.
Your coding agent learns the SearchQuery shape, importJson patterns, and ontology API — loaded when relevant, not dumped into every session.
Works with Claude Code, Cursor, Codex, Copilot, and OpenCode. No manual setup. No prompt files to maintain.
SDKs
Same API, every runtime.
All three share identical query semantics — SearchQuery, importJson, ai.search. Learn the shape once; run it in TypeScript, Python, or plain HTTP.
Agents that know your data
before they query it.
RushDB builds a live ontology from every record you push. Agents read it once per session — and never construct a query blind again.
What the agent now knows
✓ Valid filter ranges
Agent knows price ranges from $9.99 to $2,499. It won't filter for price > $10,000 and return empty results.
✓ Real property names
Agent knows the field is "category", not "type" or "productType". No hallucinated field names. No silent query failures.
✓ Traversable relationships
Agent knows CUSTOMER → ORDER → PRODUCT is the real topology. Multi-hop queries built with confidence, not guesswork.
One call. One session. No hallucinated schema.
Fewer empty results. Fewer hallucinations. Agents that explain themselves.
The same SearchQuery that finds records also discovers what labels, properties, and relationships exist — making schema introspection a natural extension of the query you’re already writing. No separate schema API to learn.
Simple, predictable pricing.
Pay for what you write — not for servers, nodes, or queries. 1 KU = 1 field written. Reads are always free. Full pricing details →
Free
Evaluate RushDB in production conditions
Get started instantly:
- 100K KU / month
~3,000 records with 10 fields each
- 2 projects
No time limits, no feature restrictions
- Full REST API and SDKs
- Self-hosted & BYOC
Connect to your own Neo4j or Aura
- Vector & AI search
- Community support
Pro
Perfect for production applications
Everything in Free, plus:
- 10M KU / month
100× Free plan — ~300,000 records with 10 fields each
- Overage billing
Continue beyond 10M KU at $3 / M KU
- Unlimited projects
- 3 team members
then $10 / seat
- Self-hosted & BYOC
Scale
For high-volume and data-intensive apps
Everything in Pro, plus:
- Unlimited KU
Usage-based — pay only for what you consume
- Unlimited team members
- Self-hosted & BYOC
- SLA guarantee
- Priority support
FAQ
Frequently asked questions