NamiDB

NamiDB is a graph database built on object storage, described as living in an S3 bucket. It combines graph, vector, and search capabilities in one engine, with an emphasis on running on storage already in use and requiring nothing to run. The product is open source.

Its core query model supports Cypher, including synchronous and asynchronous use from Python, and GQL is on the roadmap. NamiDB also includes vector search based on cosine similarity over embeddings stored as node properties, hybrid search that fuses keyword relevance and vector similarity with reciprocal rank fusion, and built-in graph algorithms such as PageRank, connected components, triangle counting, and shortest path. The page also says the engine keeps graph traversal, vector ranking, and keyword search in the same files so they can be done in one query.

NamiDB can point at an Obsidian or Markdown folder and turn it into a live graph, where notes become nodes, wikilinks become edges, and tags become a branch or tree. It ships as a Claude Code skill and exposes a hosted MCP endpoint, so agents such as Claude Code, Cursor, and other MCP clients can sync a vault and query it back with read-only tools including semantic search, backlinks, neighbors, orphans, shared tags, and raw Cypher. The product is also offered as a single Rust binary that serves a REST API and a Bolt endpoint, and as an embedded library for notebooks, scripts, local development, and CI fixtures. The page lists support for AWS S3, Cloudflare R2, Google Cloud Storage, Azure Blob, MinIO, and Tigris.

A managed cloud version is in closed beta, with namespace isolation per tenant, project, or agent. The cloud offering is described as multi-tenant and as scaling to zero when idle, with object-storage pricing based on what is stored rather than headroom. The page also says the product can be installed with pip, and that the open-source version is available through GitHub and as a Docker container.