Skip to main content
GraphCheck needs Python 3.12+ and a running Neo4j — the 5 LTS line or a documented calendar-version release (see compatibility). No account, no cloud.
No Neo4j at hand? The Docker Quickstart starts a local demo database with a seeded sample graph.

​
1. Install

The base install is lean. Optional capabilities live behind extras: pip install "graphcheck[generate]" for AI-assisted check drafting, pip install "graphcheck[mcp]" for the agent server.

​
2. Scaffold a project

In the directory that should hold your checks:
This writes three files and probes your Neo4j:
  • graphcheck.yml — project configuration
  • profiles.yml — connection details (gitignored; use password_env in CI)
  • checks/example.yml — a starter suite
Point profiles.yml at your database:

​
3. Verify the connection

This reports the server version and edition, what your credentials can and cannot see, and any blocked checks. Every error message contains its fix — see Troubleshooting for the full code list.

​
4. Run

You get a severity-weighted score, a coverage table, and two artifacts under .graphcheck/runs/latest/:
  • results.json — machine-readable verdicts, evidence, coverage, and score
  • report.html — a self-contained report that opens offline
Exit codes are a stable CI contract: 0 all executed checks passed · 1 an error-severity finding or execution error · 2 warnings or incomplete evaluation · 3 the run could not be prepared or completed.

​
Next steps