Skip to main content
The Modal Python SDK provides convenient, on-demand access to serverless cloud compute from Python scripts on your local computer.

​
Key features

Modal’s Python SDK offers powerful capabilities for running code in the cloud:
  • Serverless functions: Deploy Python functions that scale automatically
  • Container orchestration: Run code in custom container environments
  • GPU support: Access GPUs for machine learning workloads
  • Persistent storage: Use volumes, dicts, and queues for data persistence
  • Web endpoints: Create HTTP endpoints with FastAPI, ASGI, or WSGI
  • Scheduled jobs: Run functions on cron schedules or periodic intervals
  • Async support: Full support for async/await patterns

​
Python version support

The Modal SDK requires Python 3.10 through 3.14. The SDK will raise a runtime error if you attempt to use it with unsupported Python versions. From modal/__init__.py:4-7: 
if sys.version_info[:2] < (3, 10):
    raise RuntimeError("This version of Modal requires at least Python 3.10")
if sys.version_info[:2] >= (3, 15):
    raise RuntimeError("This version of Modal does not support Python 3.15+")

​
Core concepts

​
Apps

An App is the main entry point for defining Modal applications. Apps contain functions, classes, and other resources: 
import modal

app = modal.App("my-app")

@app.function()
def my_function():
    return "Hello from Modal!"

​
Functions

Functions are the primary way to run code on Modal. Decorate any Python function with @app.function() to run it in the cloud: 
@app.function()
def process_data(data: str):
    # Runs in a Modal container
    return data.upper()

​
Images

Images define the container environment for your functions. Modal provides pre-built images and supports custom configurations: 
image = modal.Image.debian_slim().pip_install("pandas", "numpy")

@app.function(image=image)
def analyze_data():
    import pandas as pd
    # Your code here

​
Resources

Modal provides several resource types for data persistence and communication:
  • Volume: Persistent file storage across function invocations
  • Dict: Key-value store for structured data
  • Queue: FIFO queue for task distribution
  • Secret: Secure storage for API keys and credentials
  • NetworkFileSystem: Shared file system for concurrent access

​
Available exports

The Modal SDK exports the following classes and functions (from modal/__init__.py:57-99): Core classes:
  • App - Main application container
  • Client - Low-level API client
  • Function, FunctionCall - Function primitives
  • Image - Container image builder
Compute resources:
  • Cls - Parameterized classes
  • Sandbox, SandboxSnapshot - Interactive environments
Storage:
  • Volume - Persistent volumes
  • Dict - Distributed dictionaries
  • Queue - Distributed queues
  • NetworkFileSystem - Network file systems
  • CloudBucketMount - Cloud storage mounting
Configuration:
  • Secret - Secure credentials
  • Retries - Retry policies
  • Cron, Period - Scheduling
  • SchedulerPlacement - Placement constraints
Decorators:
  • web_endpoint - HTTP endpoints
  • fastapi_endpoint - FastAPI integration
  • asgi_app, wsgi_app - ASGI/WSGI apps
  • batched - Batch processing
  • concurrent - Concurrent execution
Utilities:
  • is_local() - Check execution context
  • current_input_id(), current_function_call_id() - Runtime context
  • interact() - Interactive debugging
  • forward() - Port forwarding

​
Next steps

Installation

Install the Modal Python SDK with pip, uv, or poetry

Basic usage

Learn basic patterns for running functions on Modal

Async support

Use async/await with Modal functions

CLI reference

Explore Modal’s command-line interface

Build docs developers (and LLMs) love

Get started for freeTalk to us