NamiDB
v2.6.19Source-available on object storage

The bucket is the database.

NamiDB is a graph database that lives in your S3 bucket. Graph, vectors, and search in one engine, running on storage you already pay for. Nothing to run.

Get startedStar on GitHub
$pip install namidb
The NamiDB isotipo, a line-drawn portrait in Rosa NamiDB.

Starred by engineers from

  • Cloudflare
  • AT&T
  • Fly.io
  • Artium AI
  • Microsoft
  • Intel
  • Red Hat
  • Dropbox
  • Alibaba
  • Tencent
  • Weaviate
  • University of Washington

Engineers from these teams gave NamiDB a star. Add yours on GitHub.

One engine

Graph, vectors, and search.
One engine, one query.

In 2024, S3 added the last primitive it was missing, so a bucket can finally be a real database. We built the graph engine for it. Most teams glue a graph database, a vector store, and a search index together, then babysit three systems that drift apart. NamiDB keeps all three in the same files.

Cypher queries

Walk relationships with Cypher, sync or async from Python. GQL is on the roadmap.

client.cypher(...)

Vector search

Rank by meaning with cosine similarity over embeddings stored as node properties.

cosine_similarity(p.emb, $q)
One traversal

Hybrid search

Keyword relevance and vector similarity in one ranked result, fused with reciprocal rank fusion.

bm25(p.body, $kw)

Graph algorithms

PageRank, connected components, triangle counting, and shortest path, inside the engine.

CALL algo.pagerank()

Three access patterns, one set of files in your bucket. Nothing to keep in sync, because there is nothing to sync.

The product

One engine. Three deployments.
Open from day one.

v2.6.19 shipped

Server

A single Rust binary, on your bucket.

Run one Rust binary in front of your bucket. No volumes, no state to migrate, no consensus cluster to keep alive. Restart it anywhere; your data was never on the box. It serves a REST API and a Bolt endpoint, so an HTTP client, a Neo4j driver, or cypher-shell connects with nothing new to install.

  • AWS S3
  • Cloudflare R2
  • GCS
  • Azure Blob
  • MinIO
  • Tigris
$ docker run namidb-server
v2.6.19 shipped

Embedded

DuckDB for graphs.

Import the library, point it at a bucket, run Cypher inside your process. Notebooks, scripts, local dev, CI fixtures: same engine, same files. Open the same s3:// path your server uses and the graph is right there.

$ pip install namidb
Closed beta

Cloud

Namespace per tenant. Scale to zero.

Managed and multi-tenant on namidb.com, run by NamiDB, Inc. Namespace per project, per agent, per customer. Scale to zero when idle. Object-storage pricing: pay for what you store, not for headroom you don't use.

One engine, any bucket. Switch deployments or backends without a rewrite.

Obsidian + MCP

Your notes, in your agent.
A graph it can actually query.

Point NamiDB at an Obsidian or Markdown folder and it becomes a live graph: every note a node, every wikilink an edge, every tag a branch. Your agent traverses it over MCP instead of guessing.

01

Sync your vault

One command turns your Obsidian or Markdown folder into a real graph. Notes become :Note, [[wikilinks]] become LINKS_TO, #tags become a :Tag tree. Your files stay the source of truth.

terminal
# get an API key in your NamiDB dashboard, keep it in the env
export NAMIDB_API_KEY=nk_live_...
 
# sync an Obsidian or Markdown folder to a queryable graph
npx namidb-vault@latest --vault ./notes --namespace my-vault --watch
02

Point your agent at it

Add the hosted MCP endpoint to your agent config, authed by the same key. No extra server to run. Claude Code, Cursor, and any MCP client connect.

.mcp.json, hosted MCP
{
  "mcpServers": {
    "namidb": {
      "type": "http",
      "url": "https://api.namidb.com/v1/mcp",
      "headers": {
        "Authorization": "Bearer ${NAMIDB_API_KEY}",
        "X-NamiDB-Namespace": "my-vault"
      }
    }
  }
}
03

Ask it in plain English

Your agent gets read-only tools: semantic search, backlinks, neighbors, orphans, shared tags, and raw Cypher. It traverses the graph instead of guessing.

claude code
# you:
> what links to "Project X", and what's related to it?
 
# the agent calls, read-only over MCP:
backlinks(note: "Project X")
vault_search(query: "Project X")

Works with Claude Code, Cursor, and any MCP client. npx namidb-vault is open source, and namidb-mcp runs it fully local.

The terms

Pay for storage.
Read the source.

Legacy graph databases price by RAM: double your hot set, double your bill, even if your data didn't grow. NamiDB stores in object storage at object-storage prices, and ships source-available from day one under BSL 1.1.

Scale to zero

Idle namespaces cost nothing.

Compute scales to zero when nobody queries, per namespace, per tenant, per agent. You pay for the bytes at rest, not for a server keeping them warm at 3 a.m.

Backups are a copy

aws s3 sync and you're done.

Your database is just files in your bucket. Snapshots are object versions, DR is a second bucket. No proprietary dump format, no exporter to maintain, no vendor in the path.

Source on day one

BSL 1.1, then Apache 2.0.

Every architectural decision lives in an RFC and every benchmark is published, including the ones that don't go our way. Each release converts to Apache 2.0 three years later.

Your bucket, your files
Browse the sourcegithub.com/namidb/namidb
Read the researchnamidb.com/research
Cloud waitlist

The engine is here. Now we're opening the Cloud.

The engine is shipped and yours to run today. The waitlist is for NamiDB Cloud: managed and multi-tenant, a namespace per tenant, scale to zero, object-storage pricing. We onboard design partners in waves.

  • One email per wave. No spam.
  • Your data stays in your bucket. Always.
  • Unsubscribe in one click.