Operate grag

Keep a local graph available, understand failures and preserve saved knowledge.

• Share one graph

  Connect several harnesses to the same owner, or serve separate databases.

  Servers and clients →

• Find a failure

  Start from a symptom and check the selected command, database and client.

  Troubleshooting →

• Protect saved knowledge

  Export a verified snapshot and restore into a separate database.

  Backup and recovery →

• Move a checkout

  Reconcile saved paths and registrations after reorganizing folders.

  Relocation guide →

Useful first checks

grag status
grag doctor

status identifies the selected database and owner. doctor checks installation readiness. A healthy server does not establish that a particular agent's MCP connection works; see client verification.

Database cannot open?

Preserve the database, WAL and shadow files. Follow separate-copy recovery before attempting to rebuild an index.