PyMetabase

The most complete Python API wrapper and CLI for Metabase

---

Export millions of rows, manage cards, dashboards, collections, execute parameterized queries, and access every Metabase API endpoint — from Python or the command line.

🔍 Why PyMetabase?

Feature	PyMetabase	Other Libraries
Full Metabase API coverage	✅	Partial
CLI tool	✅	❌
Automatic chunking for large exports	✅	❌
Parameterized queries ({{param}})	✅	❌
Session token persistence & reuse	✅	❌
Resume interrupted exports	✅	❌
rclone-style remote management	✅	❌
Cards, Dashboards, Collections API	✅	Rare
Generic API access (any endpoint)	✅	Rare
Search, Users, Tables/Fields API	✅	❌

📦 Installation

pip install pymetabase

Optional extras:

pip install pymetabase[excel]    # Excel export (pandas + openpyxl)
pip install pymetabase[parquet]  # Parquet export (pyarrow)

⚡ Quick Start

Python — Export your first dataset in 4 lines

from pymetabase import Metabase

with Metabase(url="https://metabase.example.com", username="analyst@company.com", password="pass") as mb:
    result = mb.export(database="Production", query="SELECT * FROM customers", output="customers.jsonl")
    print(f"Exported {result.total_rows:,} rows in {result.duration_seconds:.1f}s")

CLI — Zero code, full power

# Step 1: Configure your Metabase server (one-time setup)
pymetabase config

# Step 2: Start working
pymetabase export -d Production -q "SELECT * FROM customers" -o customers.jsonl
pymetabase cards                    # see all saved questions
pymetabase search "revenue"         # find anything across your instance
pymetabase query -d Production -q "SELECT COUNT(*) FROM orders WHERE status = {{status}}" -p status=completed

---

📖 Features

🔌 Generic API Access

Hit any Metabase endpoint — no wrapper method needed:

with Metabase(url=url, username=user, password=pwd) as mb:
    # Fetch all cards
    all_cards = mb.get("/api/card")

    # Create a new question
    mb.post("/api/card", json={
        "name": "Daily Active Users",
        "dataset_query": {"database": 1, "type": "native", "native": {"query": "SELECT ..."}},
        "display": "line",
        "visualization_settings": {},
    })

    # Update and delete
    mb.put("/api/card/1", json={"name": "Weekly Active Users"})
    mb.delete_resource("/api/card/1")

# Same from the command line
pymetabase api get /api/card
pymetabase api post /api/card --json-body '{"name": "New Question", ...}'

📋 Cards (Saved Questions)

with Metabase(...) as mb:
    # Browse saved questions
    cards = mb.list_cards()
    card = mb.get_card(42)

    # Run a saved question with parameters
    result = mb.execute_card(42, parameters={"date_filter": "2024-01-01"})

    # Create a new question programmatically
    mb.create_card(
        name="Monthly Revenue by Region",
        dataset_query={
            "database": 1,
            "type": "native",
            "native": {"query": "SELECT region, SUM(amount) FROM sales GROUP BY region"},
        },
        collection_id=5,
    )

    # Housekeeping
    mb.update_card(42, name="Monthly Revenue by Region (v2)")
    mb.delete_card(99)

📊 Dashboards

with Metabase(...) as mb:
    dashboards = mb.list_dashboards()
    dash = mb.get_dashboard(1)   # includes all cards on the dashboard

    mb.create_dashboard("Q4 Executive Report", collection_id=3, description="KPIs and trends")
    mb.update_dashboard(1, name="Q4 Executive Report — Final")
    mb.delete_dashboard(1)

📁 Collections

with Metabase(...) as mb:
    collections = mb.list_collections()
    items = mb.get_collection_items("root")   # list everything at the top level
    items = mb.get_collection_items(7)         # or inside a specific collection

    mb.create_collection("Data Engineering", parent_id=1, color="#509EE3")

🔎 Search

Find cards, dashboards, collections, and tables in one call:

results = mb.search("revenue", models=["card", "dashboard"])

pymetabase search "revenue"

👤 Users

me = mb.get_current_user()               # who am I?
user = mb.get_user(1)                     # specific user
users = mb.list_users()                   # all users (admin only)

pymetabase whoami
pymetabase users

🗄️ Tables & Fields

Explore your database schema without SQL:

tables = mb.list_tables("Production")     # all tables in a database
table = mb.get_table(1)                   # table details
metadata = mb.get_table_metadata(1)       # columns, types, FKs
field = mb.get_field(42)                  # single field info
values = mb.get_field_values(42)          # distinct values (great for filters)

---

🧩 Parameterized Queries

Use {{param}} placeholders in SQL — PyMetabase handles the Metabase template-tag wiring for you:

with Metabase(...) as mb:
    # Filter orders by status and minimum amount
    rows = mb.query(
        "SELECT * FROM orders WHERE status = {{status}} AND total > {{min_total}}",
        database="Production",
        parameters={"status": "shipped", "min_total": 500},
    )

    # Export with parameters — same syntax
    mb.export(
        database="Production",
        query="SELECT * FROM signups WHERE created_at > {{since}}",
        output="recent_signups.jsonl",
        parameters={"since": "2024-01-01"},
    )

# CLI — pass parameters with -p key=value
pymetabase query -d Production \
    -q "SELECT * FROM orders WHERE status = {{status}}" \
    -p status=shipped

pymetabase export -d Production \
    -q "SELECT * FROM users WHERE role = {{role}}" \
    -o admins.jsonl -p role=admin

---

📤 Export

Large Dataset Export

Automatic chunking for datasets with millions of rows. No memory issues, no timeouts:

with Metabase(...) as mb:
    result = mb.export(
        database="Production",
        query="SELECT * FROM events",       # 10M+ rows? No problem
        output="events.jsonl",
        chunk_size=500_000,                  # fetch 500K rows at a time
        checkpoint_file="progress.json",     # resume if interrupted
        progress_callback=lambda cur, total, rate: print(f"{cur:,}/{total:,} rows ({rate:.0f}/s)"),
    )
    print(f"Done — {result.total_rows:,} rows, {result.chunks} chunks, {result.duration_seconds:.0f}s")

How it works under the hood:

1. Counts total rows with SELECT COUNT(*)
2. Wraps your query with ROW_NUMBER() OVER () and fetches in offset chunks
3. Retries failed chunks with exponential backoff
4. Saves progress to a checkpoint file so you can resume after crashes
5. Works with any SQL — CTEs, JOINs, subqueries, window functions

Export Formats

Output format is detected from the file extension:

pymetabase export -d DB -q "SELECT * FROM users" -o users.jsonl   # JSON Lines (default)
pymetabase export -d DB -q "SELECT * FROM users" -o users.json    # JSON array
pymetabase export -d DB -q "SELECT * FROM users" -o users.csv     # CSV

Table Export with Filtering

Export a table directly — no SQL needed:

with Metabase(...) as mb:
    mb.export_table(
        database="Production",
        table="users",
        output="active_users.csv",
        columns=["id", "name", "email", "plan"],
        where="active = true AND plan = 'pro'",
        order_by="created_at DESC",
        limit=10_000,
    )

---

🌐 Remote Management

Manage multiple Metabase servers like rclone manages cloud storage:

# Configure a remote (interactive)
pymetabase config

# List all configured remotes
pymetabase listremotes
# Configured remotes (2):
#   - production (default): https://metabase.company.com
#   - staging: https://staging-metabase.company.com

# Use a specific remote for any command
pymetabase -r staging export -d DB -q "SELECT 1" -o test.jsonl

# Manage remotes
pymetabase testremote production     # verify connection
pymetabase showremote production     # show config details
pymetabase selectremote staging      # change default
pymetabase deleteremote old-server   # remove a remote

# Quick identity check
pymetabase whoami

🔑 Session tokens are persisted — PyMetabase caches your auth token and reuses it across runs. No re-login on every command.

Config location:

• macOS / Linux: ~/.config/pymetabase/config.json
• Windows: %APPDATA%\pymetabase\config.json

---

⚙️ Configuration

PyMetabase supports multiple configuration methods. Highest priority wins:

1. Direct Parameters

mb = Metabase(url="https://metabase.company.com", username="analyst@company.com", password="...")

2. Remotes (recommended for CLI)

pymetabase config           # interactive setup
pymetabase -r myserver export ...

3. Credentials File

[{"SERVER_NAME": "https://metabase.company.com", "USERNAME": "analyst@company.com", "PASSWORD": "..."}]

mb = Metabase(credentials_file="credentials.json")

4. Config File (YAML)

metabase:
  url: https://metabase.company.com
  username: analyst@company.com
  password: secret

defaults:
  chunk_size: 500000
  format: jsonl

retry:
  max_retries: 3
  delay: 1.0
  timeout: 600

5. Environment Variables

export METABASE_URL=https://metabase.company.com
export METABASE_USERNAME=analyst@company.com
export METABASE_PASSWORD=secret

---

📘 CLI Reference

Command	Description
export	Export SQL query results to file
export-table	Export a table with optional column/filter selection
query	Execute SQL and print/save results (-f table|json|jsonl|csv -o file)
cards	List all saved questions
card <id>	Show card details
execute-card <id>	Run a saved question (with optional -p params)
export-card <id>	Export a saved question to file (-o file)
dashboards	List all dashboards
dashboard <id>	Show dashboard details and its cards
collections	List all collections
collection <id>	Show items inside a collection
search <query>	Search cards, dashboards, collections, tables
users	List all users (admin only)
whoami	Show current authenticated user
list-databases	List available databases
list-tables	List tables in a database
api <method> <endpoint>	Raw API request (GET/POST/PUT/DELETE)
config	Add, edit, or delete remotes
listremotes	List configured remotes
selectremote	Set the default remote
showremote <name>	Show remote configuration
deleteremote <name>	Delete a remote
testremote [name]	Test connection to a remote

---

📘 Python API Reference

Metabase Client

Metabase(
    url: str = None,
    username: str = None,
    password: str = None,
    config_file: str = None,
    credentials_file: str = None,
    remote_name: str = None,     # for token persistence
    persist_token: bool = True,  # reuse sessions across runs
)

Methods

Category	Methods
Generic API	get(endpoint) post(endpoint, **kwargs) put(endpoint, **kwargs) delete_resource(endpoint)
Cards	list_cards() get_card(id) create_card(...) update_card(id, ...) delete_card(id) execute_card(id, parameters=)
Dashboards	list_dashboards() get_dashboard(id) create_dashboard(...) update_dashboard(id, ...) delete_dashboard(id)
Collections	list_collections() get_collection(id) get_collection_items(id) create_collection(...)
Search	search(query, models=, limit=)
Users	get_current_user() list_users() get_user(id)
Tables/Fields	list_tables(db) get_table(id) get_table_metadata(id) get_field(id) get_field_values(id)
Queries	query(sql, database, parameters=)
Export	export(query, output, database, parameters=, chunk_size=, ...) export_table(table, output, ...)
Connection	connect() disconnect()

ExportResult

result.total_rows         # int — total rows exported
result.chunks             # int — number of chunks used
result.duration_seconds   # float — wall-clock time
result.output_file        # str — path to output file
result.rate_per_second    # float — rows/sec throughput
result.format             # str — "jsonl", "json", or "csv"

---

🛠️ Requirements

• Python 3.9+
• requests sqlglot pyyaml

---

🤝 Contributing

Contributions are welcome! Please open an issue or submit a pull request.

git clone https://github.com/mokarimi19/pymetabase.git
cd pymetabase
pip install -e ".[dev]"
pytest

📄 License

MIT