Relate CLI
A Cypher graph database in one binary for agents: CLI, embedded engine, MCP server and a skill for Claude Code, Cursor and Codex.
- 223 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- The skill and MCP server let an agent write to a graph database, and the install is a script from the internet
Install
- Works in
Manual install
curl -sSfL https://relatedb.ai/install.sh | bashInstalls the binary into `$HOME/.local/bin` on macOS and Linux with the checksum signature verified by the script. A script from the internet: read it first.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The skill and MCP server let an agent write to a graph database, and the install is a script from the internet
This is third-party code. Review the repository files before installing.
What it does
Relate stores a property graph and answers Cypher queries with no separate server. A graph is a single `.rl` file, a directory or in-memory only, durable through an fsync write-ahead log and Parquet checkpoints. One engine stands behind the `relate` command, a Python package, npm packages for Node.js and the browser (WASM), a Rust crate and an optional single process speaking Bolt, HTTP and MCP. There are two paths for agents. `relate skill install` drops a skill bundle with `SKILL.md` and references for the CLI, Cypher, vectors and spatial data into the directories of all detected agents, including Claude Code, Cursor and Codex. `relate serve --database ./friends --bolt 7687 --http 7474 --mcp 7475` brings up MCP over streamable HTTP for AI clients. Under a detected agent the row count in a response is capped.
Who it is for. Developers who need a graph of relations inside a project or an agent system without deploying Neo4j or another database.
Good fit when
- An agent builds a repository dependency graph and queries it in Cypher
- An assistant needs memory of links between people and tasks in one file
- A Python script needs an embedded graph database with no server process
- A browser app stores a graph on the client through WebAssembly
Not a fit when
- A production distributed database with Neo4j support is needed, since this is an experimental Neo4j Labs project
- Data does not fit in memory, since the engine keeps the graph in memory
- Full parity with Neo4j Cypher is needed, with the uncovered list in `docs/limitations.md`
Example request
Create a friends graph in `./friends`, load five people with relations and find friends of friends with a Cypher query.Limitations
This is an experimental Neo4j Labs project, not part of the supported Neo4j product lineup. The graph is held in memory, with durability from the write-ahead log and checkpoints. Some packages still build from source until the PyPI and npm publishes land, and Python wheels are on the releases page. The CLI installs through a script from the internet, and the script verifies the checksum signature itself, but it is still worth reading. The list of what the engine does not do yet is kept in `docs/limitations.md`.
How to disable. Run `relate skill remove` to uninstall the skill and delete the binary from `~/.local/bin`. Graph `.rl` files are stored separately and are deleted by hand.
FAQ
How do I install the skill for an agent?
`relate skill install` writes the skill to every detected agent and `--agent claude-code` limits it to one. `relate skill remove` removes it.
How do I start MCP?
With `relate serve --database ./friends --mcp 7475`, which brings up MCP over streamable HTTP, while `--bolt` and `--http` add the other interfaces.
Where is the graph stored?
In a single `.rl` file, a directory or in-memory only. Durability comes from an fsync write-ahead log and Parquet checkpoints.
Which install options exist?
The `curl -sSfL https://relatedb.ai/install.sh | bash` script for the CLI, `pip install relate-db`, npm packages and a Rust crate.
Can the CLI be updated?
Yes. `relate upgrade` downloads, verifies and replaces the binary, and `relate upgrade --check` only reports the version.
Discussion
No questions or reviews yet.
Related
CLI toolsA single binary indexes tables and logs so an agent gets a few matching records back instead of raw history
KGLite keeps a knowledge graph in one file, and its bundled MCP server gives agents Cypher queries with no separate database.
Neo4j skills for agents: Cypher queries, graph modeling, drivers, import, GraphRAG, GDS and Aura provisioning
A graph database with git-style branches: the server exposes it to agents over MCP at /mcp, and a skill teaches the agent schemas and queries
A CLI and MCP server that extracts structured knowledge from documents: graphs, hypergraphs, and fact search with sources
A local MCP server that turns a folder of documents into a knowledge base with hybrid search for Claude Code and Cursor
README in short
One engine is available through the CLI, libraries and a server process. A skill and MCP are added for agents, and response size is capped under an agent.