Quickstart
Start a local development database and Explorer using the images on Docker Hub:
ianknowles/galactus-db:latest (currently 2.5.0) and
ianknowles/galactus-db-explorer:latest. You need Docker with Compose.
Both images support Linux ARM64 (aarch64) and Linux AMD64 (x86-64). Docker selects the image for your container host; see supported architectures.
Start the database and Explorer
- Save the Compose configuration
as
compose.yamlin a new directory. - Create a
.envfile beside it and choose a local database password of at least 12 characters:
GDB_INITIAL_PASSWORD=replace-with-your-password
- In that directory, start the services:
docker compose up -d
docker compose ps
The example runs the free Community edition with two allocated CPUs, persistent storage and group commit durability. Community is free for any use, including production, and needs no licence key or website account. Commercial adds any number of databases and licensed backup and restore. See editions and licences.
With the MCP server for AI agents
To connect AI assistants and agents as well, use the
Compose configuration with the MCP server
in step 1 instead. It starts the database, Explorer and
ianknowles/galactus-db-mcp:latest with the same .env file and commands.
See Connect an AI agent once it is running.
Open Explorer
Visit http://127.0.0.1:7474. The connection is prefilled with
bolt://db:7687, the database's address inside the Compose network.
Log in with username gdb and the password from your .env file.
Select the galactus database. Applications running on your computer connect
to bolt://127.0.0.1:7687; applications in the same Compose network use
bolt://db:7687.
The username is gdb. Release images require a password of at least 12 characters
before they start; set it explicitly in your .env file.
Create and query a graph
Paste this script into Explorer and run it. Each semicolon-delimited statement commits separately.
CREATE (a:Person {name: 'Ada', age: 36})-[:KNOWS {since: 2015}]->(b:Person {name: 'Bob'});
MATCH (a:Person)-[r:KNOWS]->(b:Person)
WHERE a.age >= 30
RETURN a.name AS who, b.name AS knows, r.since
ORDER BY who;
CREATE INDEX ON :Person(name);
MATCH (p:Person {name: 'Ada'}) RETURN p.name, p.age;
The result is Ada knowing Bob, with since equal to 2015. Rerunning CREATE
adds another pair; use MERGE for get-or-create behaviour.
Connect an AI agent
If you started the stack with the MCP server, MCP clients connect to
http://127.0.0.1:7688/mcp and sign in with a Galactus DB user and password
over HTTP Basic authentication. Each user sees the tools their roles allow.
Give an agent its own user rather than gdb. In Explorer, run:
CREATE USER agent SET PASSWORD 'choose-a-password-for-the-agent' CHANGE NOT REQUIRED;
GRANT ROLE reader TO agent;
With the reader role, the agent can read the schema, run read-only Cypher and
search the documentation, but cannot change data. Then add the server to your
client. For Claude Code:
claude mcp add --transport http galactus-db http://127.0.0.1:7688/mcp --header "Authorization: Basic $(printf 'agent:choose-a-password-for-the-agent' | base64)"
See the MCP server guide for desktop clients, the tools and security.
Stop and restart
docker compose down
docker compose up -d
Your named data volume survives these commands. Removing the volume deletes its data. See Docker deployment for upgrades and configuration.
Related articles
Connect a native driver · Driver repositories · Explorer · MCP server · Cypher