Skip to main content
Podflare is a cloud sandbox platform built for AI agents. Every sandbox runs in its own Firecracker microVM — hardware-isolated, not a container — with a persistent Python REPL and a copy-on-write filesystem. Your agent can create a sandbox in 7–11 ms from the warm pool, fork it N ways to explore multiple plans in parallel, and commit the winner back into the parent.

Quick Start

Install the SDK and run your first code in minutes

API Reference

Full HTTP API for creating, executing, and managing sandboxes

Python SDK

Complete Python SDK reference with examples

TypeScript SDK

Full TypeScript SDK with async/await and type annotations

​
What you can do with Podflare

Fork sandboxes

Snapshot a running sandbox and branch N isolated copies in ~101 ms

Persistent REPL

Variables, imports, and state survive across multiple run_code calls

Integrate with AI frameworks

Drop-in tools for OpenAI Agents SDK, Anthropic, and Vercel AI SDK

MCP server

Add Podflare to Claude Desktop, Cursor, Cline, or Zed via MCP

​
Get started in 3 steps

1

Install the SDK

Install the Python or TypeScript SDK from PyPI or npm.
pip install podflare
# or
npm install podflare
2

Set your API key

Export your API key so the SDK can authenticate with Podflare.
export PODFLARE_API_KEY=pk_your_key_here
3

Run your first code

Create a sandbox, run code, and close it.
from podflare import Sandbox

with Sandbox() as s:
    r = s.run_code("print(sum(range(10)))")
    print(r.stdout)  # 45