GitHub - andreagrandi/mb-cli: CLI client for MetaBase

A read-only CLI for the Metabase API. Designed for terminal users and AI coding agents (Claude, Codex) to explore databases, inspect schemas, and run ad-hoc queries.

Installation

Homebrew

brew tap andreagrandi/tap
brew install mb-cli

Binary download

Download the latest release from the releases page.

Go install

go install github.com/andreagrandi/mb-cli/cmd/mb@latest

Configuration

Set two environment variables:

export MB_HOST=https://your-metabase-instance.com
export MB_API_KEY=your-api-key

Both are required. MB_HOST is the base URL of your Metabase instance. MB_API_KEY is a Metabase API key.

Optional:

export MB_REDACT_PII=false  # Disable PII redaction (enabled by default)

Usage

Global flags

Flag	Description	Default
`--format`, `-f`	Output format: `json`, `table`	`json`
`--verbose`, `-v`	Show request details on stderr	`false`
`--redact-pii`	Redact PII values in query results	`true`

Database commands

# List all databases
mb-cli database list

# Get database details
mb-cli database get 1

# Full metadata (tables + fields)
mb-cli database metadata 1

# List all fields in a database
mb-cli database fields 1

# List schema names
mb-cli database schemas 1

# List tables in a specific schema
mb-cli database schema 1 public

The database command has an alias db:

Table commands

# List all tables
mb-cli table list

# Get table details
mb-cli table get 42

# Table metadata with field details
mb-cli table metadata 42

# Foreign key relationships
mb-cli table fks 42

# Raw table data
mb-cli table data 42

Field commands

# Get field details
mb-cli field get 100

# Summary statistics
mb-cli field summary 100

# Distinct values
mb-cli field values 100

SQL queries

# Run a SQL query by database ID
mb-cli query sql --db 1 --sql "SELECT * FROM users LIMIT 10"

# Run a SQL query by database name (case-insensitive substring match)
mb-cli query sql --db prod --sql "SELECT count(*) FROM orders"

# Append a LIMIT clause
mb-cli query sql --db 1 --sql "SELECT * FROM users" --limit 5

# Export results as CSV
mb-cli query sql --db 1 --sql "SELECT * FROM users" --export csv

The --db flag accepts a numeric database ID or a name substring. If the name matches multiple databases, you'll be asked to use the ID instead.

Saved questions (cards)

# List all saved questions
mb-cli card list

# Get card details
mb-cli card get 10

# Get the full card definition, including dataset_query and template tags
mb-cli card get 10 --full

# Execute a saved question
mb-cli card run 10

# Execute a saved question with parameters
mb-cli card run 10 --param timeframe_days=14

Dashboards

# List all dashboards
mb-cli dashboard list

# Get dashboard metadata, tabs, filters, and grouped cards
mb-cli dashboard get 298

# List the saved questions used by a dashboard
mb-cli dashboard cards 298

# Discover valid values for a dashboard filter
mb-cli dashboard params values 298 merchant_name

# Search dashboard filter values
mb-cli dashboard params search 298 merchant_name acme

# Execute a dashboard card with dashboard parameters applied
mb-cli dashboard run-card 298 1201 398 --param merchant_name="Acme Corp"

# Summarize tabs, parameter mappings, and source-card dependencies
mb-cli dashboard analyze 298

Search

# Search across all Metabase items
mb-cli search "users"

# Filter by type
mb-cli search "revenue" --models table,card

Version

Agent integration

mb-cli is designed to be used by AI coding agents. When piped or used in scripts, output defaults to JSON for easy parsing.

Typical agent workflow:

# 1. Discover databases
mb-cli database list

# 2. Find relevant tables
mb-cli search "users" --models table

# 3. Inspect table schema
mb-cli table metadata 42

# 4. Query data
mb-cli query sql --db 1 --sql "SELECT id, email FROM users WHERE created_at > '2024-01-01' LIMIT 10"

# 5. Inspect a dashboard that depends on saved questions
mb-cli dashboard analyze 298

Dashboard analysis workflow

Example flow for dashboard 298:

# 1. Inspect dashboard structure
mb-cli dashboard get 298

# 2. List the saved questions behind the dashboard
mb-cli dashboard cards 298

# 3. Inspect a card's full SQL or MBQL definition
mb-cli card get 398 --full

# 4. Discover valid parameter values
mb-cli dashboard params values 298 merchant_name

# 5. Summarize dependencies and assumption-backed cards
mb-cli dashboard analyze 298

API coverage

The dashboard and parameter workflows use these Metabase endpoints:

Command	Endpoint
`dashboard list`	`GET /api/dashboard/`
`dashboard get`	`GET /api/dashboard/:id`
`dashboard cards`	`GET /api/dashboard/:id`
`dashboard params values`	`GET /api/dashboard/:id/params/:param-key/values`
`dashboard params search`	`GET /api/dashboard/:id/params/:param-key/search/:query`
`dashboard run-card`	`POST /api/dashboard/:dashboard-id/dashcard/:dashcard-id/card/:card-id/query`
`card get --full`	`GET /api/card/:id`
`card run --param`	`POST /api/card/:card-id/query`
`dashboard analyze`	`GET /api/dashboard/:id`, `GET /api/card/:id`

PII Redaction

When AI agents use mb-cli directly (via shell commands), query results containing PII (emails, names, phone numbers) flow through stdout into the model's context. This feature prevents agents from seeing sensitive data by redacting PII columns before data leaves the client layer. Agents can still cross-reference records using IDs; for actual PII values, the user can check directly in Metabase.

The approach leverages Metabase's own field semantic types (type/Email, type/Name, etc.). Redaction is ON by default — no configuration needed. Users must explicitly opt out with MB_REDACT_PII=false or --redact-pii=false. This is defense-in-depth: it makes the default path safe. An agent would have to actively choose to bypass it (a visible, auditable action).

License

MIT