The memory and storage engine for AI agents — from the edge to the cloud.
Durable, queryable, access-guarded memory for AI agents that runs where the agent runs —
embedded on device, offline-capable, and syncing to a distributed server when connected.
Semantic, relational, episodic, and time-series memory in one engine, one process, with no
network hops between them.
Quickstart
·
Docs
·
Benchmarks
·
NodeDB-Lite
·
CLI
AI agents need memory: what they were told, how facts relate, what happened when, and what matters right now. The usual answer stitches a cloud vector store to a graph database to a document store — or dumps everything into a markdown file and hopes. NodeDB puts all of it in one engine that runs where the agent runs: the same engine embeds on a laptop, phone, or browser ([NodeDB-Lite](https://github.com/NodeDB-Lab/nodedb-lite)) and scales to a distributed cluster (NodeDB Origin), with CRDT sync between them.
It already powers **ma8e**, an in-development memory-and-knowledge layer for coding agents built entirely on embedded NodeDB — deterministic capture, hybrid retrieval, and access guards in one process with no network hop.
## Why NodeDB
- **Memory that runs where the agent runs.** The same engines embed on device and in the browser (NodeDB-Lite, WASM), work offline, and sync conflict-free to the server (CRDT). A cloud vector store can't embed next to your agent; a markdown file isn't queryable or access-guarded.
- **All of an agent's memory in one query.** Vector search, graph traversal, full-text ranking, document storage, timeseries, and key-value share one storage core, one identity space, and one planner — so a query flows from semantic search into graph traversal filtered by full-text, in one process with no network hops. Multi-model is _how_ NodeDB does this; the point is the capability, not the label.
- **PostgreSQL wire protocol.** Connect with `psql` or any PostgreSQL client library. Standard SQL with engine-specific extensions where SQL can't express the operation.
- **One engine instead of a polyglot stack.** When you outgrow the edge, the same engine folds a server-side sprawl of separate systems into a single deployment — fewer services to run, no cross-system sync pipelines, no schema drift.
- **Serious about performance.** Thread-per-Core data plane with io_uring, SIMD-accelerated distance functions, zero-copy MessagePack transport, per-column compression (ALP, FastLanes, FSST, Gorilla). See benchmarks below.
## Performance
**Timeseries ingest + query benchmark** — 10M rows, high-cardinality DNS telemetry (50K+ unique domain names). Single node, NVMe storage.
### Ingest
| Engine | Rate | Time | Memory | Disk |
| ----------- | ------------ | ------------ | ---------- | ------------ |
| **NodeDB** | **93,450/s** | 107s | **120 MB** | 2,217 MB |
| TimescaleDB | 56,615/s | 177s | 963 MB | 2,802 MB |
| ClickHouse | 53,905/s | 186s | 1,035 MB | **1,647 MB** |
| InfluxDB | 22,715/s | 88s (2M cap) | 1,656 MB | 982 MB |
### Queries (ms, best of 3)
| Query | NodeDB | ClickHouse | TimescaleDB | InfluxDB (2M) |
| -------------------------- | ------- | ---------- | ----------- | ------------- |
| `COUNT(*)` | **<1** | 1 | 423 | 13,110 |
| `WHERE qtype=A COUNT` | 47 | **6** | 347 | 5,297 |
| `WHERE rcode=SERVFAIL` | 41 | **6** | 334 | 1,048 |
| `GROUP BY qtype` | 56 | **15** | 597 | 12,426 |
| `GROUP BY rcode` | 52 | **16** | 604 | 13,183 |
| `GROUP BY cached+AVG` | 120 | **33** | 677 | 13,652 |
| `GROUP BY client_ip (10K)` | **141** | 157 | 660 | 14,301 |
| `GROUP BY qname (50K+)` | 2,665 | **288** | 3,644 | 16,720 |
| `time_bucket 1h` | 101 | **30** | 603 | -- |
| `time_bucket 5m+qtype` | 138 | **99** | 711 | -- |
NodeDB is not a specialized timeseries database, yet it ingests 1.65x faster than TimescaleDB and 1.73x faster than ClickHouse with 8x less memory. Query latency is competitive with ClickHouse on low-cardinality aggregations and within 3-5x on high-cardinality GROUP BY. This is the tradeoff of a general-purpose engine: you get one system instead of five, with performance that stays in the same ballpark as specialized tools.
## Engines
| Engine | What it replaces | Key capability |
| -------------------------------------------- | ---------------------------- | ----------------------------------------------------------------------------------------------------------------------- |
| [Vector](docs/vectors.md) | pgvector, Pinecone, Weaviate | HNSW with SQ8/PQ quantization, adaptive bitmap pre-filtering |
| [Graph](docs/graph.md) | Neo4j, Amazon Neptune | CSR adjacency, 13 algorithms, Cypher-subset MATCH, GraphRAG |
| [Document](docs/documents.md) | MongoDB, CouchDB | Schemaless (MessagePack + CRDT) or Strict (Binary Tuples, O(1) field access). Typeguards for gradual schema enforcement |
| [Columnar](docs/columnar.md) | ClickHouse, DuckDB | Per-column codecs (ALP, FastLanes, FSST), predicate pushdown, HTAP bridge |
| [Timeseries](docs/timeseries.md) | TimescaleDB, InfluxDB | ILP ingest, continuous aggregation, PromQL, approximate aggregation |
| [Spatial](docs/spatial.md) | PostGIS | R\*-tree, geohash, H3, OGC predicates, hybrid spatial-vector |
| [Key-Value](docs/kv.md) | Redis, DynamoDB | O(1) lookups, TTL, sorted indexes, rate limiting, SQL-queryable |
| [Full-Text Search](docs/full-text-search.md) | Elasticsearch | BMW BM25, 27-language support, CJK bigrams, fuzzy, hybrid vector fusion |
| [Array (NDArray)](docs/array.md) | TileDB, Zarr | Multi-dimensional tiles, Z-order indexing, bitemporal support, tile-level retention |
## Install
```bash
# Docker
docker run -d \
-p 6432:6432 -p 6433:6433 -p 6480:6480 \
-v nodedb-data:/var/lib/nodedb \
farhansyah/nodedb:latest
# Cargo
cargo install nodedb
```
Requires Linux kernel >= 5.1 (io_uring). Connect:
```bash
ndb # native CLI (connects to localhost:6433)
psql -h localhost -p 6432 # or any PostgreSQL client
```
```sql
CREATE COLLECTION users;
-- Standard SQL
INSERT INTO users (name, email, age) VALUES ('Alice', 'alice@example.com', 30);
-- Object literal syntax (same result)
INSERT INTO users { name: 'Bob', email: 'bob@example.com', age: 25 };
-- Batch insert
INSERT INTO users [
{ name: 'Charlie', email: 'charlie@example.com', age: 35 },
{ name: 'Dana', email: 'dana@example.com', age: 28 }
];
SELECT * FROM users WHERE age > 25;
```
New here? Start with the [Quickstart](https://nodedb.dev/docs/introduction/quickstart/) on the official documentation site: **[nodedb.dev/docs](https://nodedb.dev/docs)**.
## Deployment Modes
| Mode | Use case |
| ------------------- | ---------------------------------------------------------------------------- |
| **Origin (server)** | Full distributed database. Multi-Raft, io_uring, pgwire. Horizontal scaling. |
| **Origin (local)** | Same binary, single-node. No cluster overhead. |
| **NodeDB-Lite** | Embedded library for phones, browsers, desktops. CRDT sync to Origin. |
## NodeDB-Lite
All eight engines as an embedded library. Linux, macOS, Windows, Android, iOS, and browser (WASM, experimental).
- **Lite only** -- local-first apps that don't need a server. Vector search, graph, FTS, documents, arrays, all in-process with sub-ms reads.
- **Lite + Origin** -- offline-first with CRDT sync. Writes happen locally, deltas merge to Origin when online. Multiple devices converge regardless of order.
- **Same API** -- the `NodeDb` trait is identical across Lite and Origin. Switch between embedded and server without changing application code.
See [NodeDB-Lite](https://github.com/NodeDB-Lab/nodedb-lite) for platform details and sync configuration.
## Key Features
**Write-time validation** -- Typeguards enforce types, required fields, CHECK constraints, and DEFAULT/VALUE expressions on schemaless collections. Graduate to strict schema with `CONVERT COLLECTION x TO document_strict`.
**Bitemporal queries** -- System time (audit trail) and valid time (temporal semantics) across all engines. Query data as it existed in the past, or as it was valid on a past date. GDPR-compliant tile purge on array engine.
**Multi-dimensional arrays** -- Scientific computing with Z-order indexed tiles, tile-level compression, and bitemporal support. Combine with vector/graph/text in fused queries via cross-engine identity.
**Real-time** -- CDC change streams with consumer groups (~1-5ms latency). Streaming materialized views. Durable topics. Cron scheduler. LISTEN/NOTIFY. All powered by the Event Plane.
**Programmability** -- Stored procedures with `IF/FOR/WHILE/LOOP`. User-defined functions. Triggers (async, sync, deferred). `SECURITY DEFINER`.
**Security** -- RBAC with GRANT/REVOKE. Row-level security with `$auth.*` context across all engines. Hash-chained audit log. Multi-tenancy with per-tenant encryption. JWKS, mTLS, API keys.
**Six wire protocols** -- pgwire (PostgreSQL), HTTP/REST, WebSocket, RESP (Redis), ILP (InfluxDB line protocol), native MessagePack.
## Tools
- **[`ndb`](https://github.com/NodeDB-Lab/nodedb-cli)** -- Native CLI with TUI, syntax highlighting, and tab completion. Alternative to `psql`.
- **[NodeDB Studio](https://github.com/NodeDB-Lab/nodedb-studio)** -- GUI client for managing collections, browsing data, and monitoring. _(coming soon)_
- **[nodedb-bench](https://github.com/NodeDB-Lab/nodedb-bench)** -- Performance benchmarks against competing databases.
## Documentation
The official documentation site is **[nodedb.dev/docs](https://nodedb.dev/docs)** — start with the [Quickstart](https://nodedb.dev/docs/introduction/quickstart/).
In-repo references:
- [Getting Started](docs/getting-started.md) -- Build, run, connect
- [Architecture](docs/architecture.md) -- Three-plane execution model
- [Engine Guides](docs/README.md) -- Deep dives into each engine
- [Security](docs/security/README.md) -- Auth, RBAC, RLS, audit, multi-tenancy
- [Real-Time](docs/real-time.md) -- CDC, pub/sub, LIVE SELECT
- [NodeDB-Lite](https://github.com/NodeDB-Lab/nodedb-lite) -- Embedded edge database
- [AI Patterns](docs/ai/README.md) -- RAG, GraphRAG, agent memory, feature store
## Contributing
We welcome bug fixes, engine improvements, new codecs and analyzers, test coverage, and documentation. Read [CONTRIBUTING.md](CONTRIBUTING.md) before opening a PR — NodeDB's Three-Plane execution model has hard rules that reviewers enforce.
## Building from Source
For development or contributing:
```bash
git clone https://github.com/NodeDB-Lab/nodedb.git
cd nodedb
cargo build --release
cargo install cargo-nextest --locked # one-time
cargo nextest run --all-features
```
## Release Status
NodeDB Origin is in **public beta**; the current release is **v0.5.0 (2026-08-04)**. All eight engines are feature-complete and covered by tests. The query-facing wire protocols (pgwire, HTTP, native MessagePack, RESP) are stable — clients written against 0.1.0 will keep working through 1.0. The CRDT sync protocol and ILP ingest are still moving: 0.5.0 changes the sync frame and requires authentication on ILP connections, so upgrade NodeDB Lite and ingest clients together with the server.
v0.5.0 ships as server binaries and Docker images; the crates.io publication follows separately, so `cargo add` still resolves 0.4.0 for now. Library users who need the 0.5.0 fixes today can depend on the `v0.5.0` tag by git.
**v0.5.0 — Beta (today).** Build new products on it. The public surface (SQL dialect, query wire protocols, configuration) is stable; expect internal changes (storage layout, on-disk format, replication internals) between minor releases — a 0.4.0 data directory needs a dump/reload to move to 0.5.0. Patch and minor bumps land as needed — 0.2 through 0.5 shipped as the work demanded. The beta period is deliberately for real workloads to surface edge cases we can't manufacture in-house.
**v1.0.0 — Production-ready (target: after the beta hardening period).** What 1.0 guarantees:
- **API & SQL stability** — semver from 1.0 onward. No breaking SQL or client-API changes within a major.
- **Wire protocol stability** — pgwire, HTTP, native MessagePack, RESP, ILP, WebSocket frozen.
- **On-disk format stability** — no breaking migrations within 1.x. Forward-compatible upgrades only.
- **Cluster & Raft stability** — rolling upgrades supported within 1.x; no quorum-breaking changes.
- **Performance SLAs** — published p50/p99 targets per engine, regression-gated in CI.
- **Security audit** — third-party audit completed and findings remediated before 1.0 ships.
- **Storage, backup, and recovery** — fully exercised under fault injection, sustained load, and crash-restart cycles.
Pre-1.0 versions may change internals between releases — those changes are critical-path work (storage, backup, security, recovery) that has to be hardened in real production conditions before we put a stability stamp on it. The wire protocol and SQL surface won't break; everything underneath is fair game until 1.0.
> **Note:** This release track applies to **NodeDB Origin** (the server) only. [NodeDB-Lite](https://github.com/NodeDB-Lab/nodedb-lite), [`ndb` CLI](https://github.com/NodeDB-Lab/nodedb-cli), and [NodeDB Studio](https://github.com/NodeDB-Lab/nodedb-studio) are versioned independently on their own tracks.
**Want to test or experiment with NodeDB?** Join our [Discord](https://discord.gg/s54gDMVc7B) — we provide full support for early adopters during the beta.
## Contributors