CLI Reference

Run dinobase --help for a summary, or dinobase <command> --help for any command.

`dinobase init`

Create the config directory and database.

dinobase init [OPTIONS]

Option	Description
`--storage`	Cloud storage URL (e.g., `s3://bucket/dinobase/`, `gs://bucket/dinobase/`, `az://container/dinobase/`)

Creates ~/.dinobase/ with config.yaml and dinobase.duckdb. Safe to run multiple times.

With --storage, data is stored in cloud storage instead of locally. See the Cloud Storage Backend guide.

dinobase init                                    # local (default)
dinobase init --storage s3://bucket/dinobase/    # S3
dinobase init --storage gs://bucket/dinobase/    # GCS
dinobase init --storage az://container/dinobase/ # Azure

`dinobase add <type>`

Add a data source.

dinobase add <source_type> [OPTIONS]

Option	Description
`--name`	Custom name for the source (defaults to type)
`--path`	Path to files (parquet/csv sources only)
`--sync-interval`	Sync interval (e.g., `30m`, `1h`, `6h`)
`--freshness`	Freshness threshold (e.g., `1h`, `30m`). Defaults: 1h for SaaS, 6h for databases

Source-specific flags (e.g., --api-key, --connection-string) are passed through.

dinobase add stripe --api-key sk_live_...
dinobase add stripe --api-key sk_live_... --freshness 30m
dinobase add parquet --path ./data/ --name analytics
dinobase add postgres --connection-string postgresql://... --name prod --sync-interval 30m

`dinobase sources`

List connected data sources, or all available source types.

dinobase sources [OPTIONS]

Option	Description
`--available`	Show all available source types (not just connected)
`--pretty`	Human-readable output

By default shows only your connected sources. Use --available to see all 100+ supported source types grouped by category.

`dinobase sync [source]`

Sync data from connected sources.

dinobase sync [SOURCE_NAME] [OPTIONS]

Option	Default	Description
`--schedule`	off	Run as daemon
`--interval`	`1h`	Default interval for `--schedule`
`--max-workers`	`10`	Max concurrent syncs

dinobase sync                              # all sources, once
dinobase sync stripe                       # one source
dinobase sync --schedule --interval 30m    # daemon mode

`dinobase refresh [source]`

Re-sync sources to get fresh data.

dinobase refresh [SOURCE_NAME] [OPTIONS]

Option	Description
`--stale`	Refresh only sources that exceed their freshness threshold
`--pretty`	Human-readable output

dinobase refresh                 # refresh all sources
dinobase refresh stripe          # refresh one source
dinobase refresh --stale         # refresh only stale sources
dinobase refresh --stale --pretty

Without arguments, refreshes all non-file sources. Use --stale to only refresh sources past their freshness threshold.

`dinobase status`

Show status of all sources.

dinobase status [OPTIONS]

Option	Description
`--pretty`	Human-readable output

`dinobase query "<sql>"`

Execute a SQL query (DuckDB dialect).

dinobase query "<sql>" [OPTIONS]

Option	Default	Description
`--pretty`	off	Table output
`--max-rows`	`200`	Max rows returned

dinobase query "SELECT * FROM stripe.customers LIMIT 5" --pretty
dinobase query "SELECT COUNT(*) FROM hubspot.contacts"

`dinobase describe <table>`

Describe a table’s columns, types, annotations, and sample data.

dinobase describe <table> [OPTIONS]

Option	Description
`--pretty`	Human-readable output

table can be schema.table or just table (searches all schemas).

`dinobase info`

Show database overview for agents. Outputs the same text used as MCP server instructions.

dinobase info

No options. Always plain text.

`dinobase serve`

Start the MCP server (stdio transport).

dinobase serve [OPTIONS]

Option	Default	Description
`--sync`	off	Enable background sync
`--sync-interval`	`1h`	Sync interval

dinobase serve                              # MCP server only
dinobase serve --sync --sync-interval 30m   # with background sync

`dinobase confirm <mutation_id>`

Confirm and execute a pending mutation.

dinobase confirm <mutation_id>

Mutations submitted via dinobase query "UPDATE ..." return a preview with a mutation_id. Use this command to execute it.

dinobase query "UPDATE stripe.customers SET name = 'Acme' WHERE id = 'cus_123'"
# Returns preview with mutation_id
dinobase confirm mut_abc123def456

`dinobase cancel <mutation_id>`

Cancel a pending mutation without executing it.

dinobase cancel <mutation_id>

`dinobase auth <type>`

Connect a source via OAuth (browser-based authorization).

dinobase auth <source_type> [OPTIONS]

Option	Description
`--name`	Custom name for the source (defaults to type)
`--proxy-url`	OAuth proxy URL (or set `DINOBASE_OAUTH_PROXY_URL`)

Opens your browser to authorize Dinobase to access the source. Tokens are stored locally and refreshed automatically on sync.

dinobase auth hubspot
dinobase auth salesforce --name my_salesforce

`dinobase mcp-config [client]`

Print MCP configuration for Claude Code, Claude Desktop, or Cursor.

dinobase mcp-config [CLIENT]

Argument	Description
`claude-desktop`	Config for `~/.claude/claude_desktop_config.json`
`claude-code`	Config for `.mcp.json` (project root)
`cursor`	Config for `.cursor/mcp.json` (project root)

Without arguments, prints configs for all three clients.

dinobase mcp-config                # show all configs
dinobase mcp-config claude-desktop # Claude Desktop only
dinobase mcp-config claude-code    # Claude Code only
dinobase mcp-config cursor         # Cursor only

Output format

All data commands (status, query, describe) output JSON by default, optimized for agent consumption. Add --pretty for human-readable tables.