# Database Availability and Platform

When a registered database becomes searchable and which filesystems KuraDB supports.

## Database availability

Registry changes are durable immediately, but the daemon constructs its loaded database map only during startup. After `kura add`, restart KuraDB before sending searches to the new name. `/api/list` exposes `registered` and `loaded` separately so clients can detect this state.

A registered database is skipped at startup, and stays out of `loaded`, when its directory or inbox cannot be created, its `~/Kura_{name}` path exists but is not a symlink, or its `data.db` fails to open. Each skip is logged to `daemon.log`.

`kura mcp` builds its own loaded map when it is spawned. Restart the MCP session to see databases added after that point; see [CLI: MCP](/cli-reference-mcp).

## Platform constraints

Use a local APFS/HFS+ filesystem on macOS or ext4/xfs on Linux. The watcher depends on reliable POSIX directory mtime semantics and does not support Windows, SMB, NFS, or FUSE mounts.
