An agent that can create databases will create several, most of them scratch. Having them local by default means that costs nothing and leaks nothing, and promoting one to synced when it turns out to matter is a decision rather than a default. Querying sorted by a field rather than only by id is the minimum that makes a document store useful for anything beyond key-value, and it is here. Small surface, honest scope: this is a place for an agent to keep structured state between sessions, not a database you would build a product on.
A Fireproof-backed JSON document store where databases are created and managed as objects, with optional cloud sync per database.
- Databases created, listed and deleted as first-class things, so an agent can partition its own storage rather than sharing one bucket.
- Documents saved, loaded by id, queried sorted by a field, and deleted.
- connect_json_doc_database_to_cloud attaches a database to cloud sync and shows the dashboard, so local-first becomes shared when you want it to.
- list_json_doc_databases is described as the way to understand which databases exist before working against one.
Nothing to supply for local use. Cloud sync is opt-in per database rather than the default.
One command — npx -y @jimpick/mcp-json-db-collection-server