# Podflare

## Docs

- [Create a sandbox](https://docs.podflare.ai/api-reference/create-sandbox.md): Returns a sandbox id drawn from the warm pool (7–11 ms) or cold-booted if the pool is empty.
- [Destroy a sandbox](https://docs.podflare.ai/api-reference/destroy.md): Destroy the sandbox and free its resources.
- [Execute code](https://docs.podflare.ai/api-reference/exec.md): Run code in the sandbox. Streams NDJSON events as they happen.
- [Fork](https://docs.podflare.ai/api-reference/fork.md): Snapshot a running sandbox and spawn N copies from that state.
- [API Reference](https://docs.podflare.ai/api-reference/introduction.md): HTTP endpoints on the hostd. Same contract Python + TypeScript SDKs speak.
- [Merge into](https://docs.podflare.ai/api-reference/merge-into.md): Commit a fork child as the new state of its parent.
- [vs E2B, Daytona, and Blaxel](https://docs.podflare.ai/architecture/comparison.md): Head-to-head benchmarks against the three other major AI-agent sandbox platforms. Same harness, same machine, same minute.
- [Performance](https://docs.podflare.ai/architecture/performance.md): Real end-to-end numbers. Every ms is measured, not modeled.
- [Authentication](https://docs.podflare.ai/concepts/auth.md): Bearer-token API auth. Clerk-backed control plane comes when we ship the dashboard.
- [Egress policy](https://docs.podflare.ai/concepts/egress.md): How Podflare sandboxes reach the outside network — and how to lock that down when you don't want them to.
- [Fork](https://docs.podflare.ai/concepts/fork.md): Snapshot a running sandbox, spawn N copies, each starts from the parent's current state.
- [Observability](https://docs.podflare.ai/concepts/observability.md): W3C Trace Context propagation, structured spans, JSON logs.
- [Persistent REPL](https://docs.podflare.ai/concepts/repl.md): Variables, imports, and in-memory state survive across run_code calls.
- [What is a Podflare sandbox?](https://docs.podflare.ai/concepts/sandboxes.md): One sandbox is a full hardware-isolated Linux microVM with network, filesystem, and a live Python REPL. Engineer-grade. Real capabilities.
- [Warm Pool](https://docs.podflare.ai/concepts/warm-pool.md): How create() returns in 6 ms server-side, 187 ms end-to-end.
- [Code-interpreter agents](https://docs.podflare.ai/examples/code-interpreter.md): Wire Podflare sandboxes into Claude, OpenAI Agents, or Vercel AI SDK as the execution backend. One helper per framework, same hardware-isolated microVM underneath.
- [Database work](https://docs.podflare.ai/examples/database.md): Ephemeral Postgres per agent session. Local Supabase for testing generated RLS policies, migrations, and RESTful queries against a fresh DB.
- [Dev environments](https://docs.podflare.ai/examples/dev-environments.md): Clone a git branch, run tests, return results. The pattern behind PR-review bots and 'test my code' agents.
- [Examples](https://docs.podflare.ai/examples/overview.md): Copy-paste patterns for real AI-agent workflows: code interpreters, data analysis, ephemeral databases, browser automation, dev-env previews.
- [Web automation](https://docs.podflare.ai/examples/web-automation.md): Playwright-driven browser automation in an isolated Chromium. Boot a Next.js dev server and screenshot agent-generated websites.
- [Anthropic Messages API](https://docs.podflare.ai/integrations/anthropic.md): Handle `code_execution_20250825` tool_use blocks with Podflare.
- [Claude Desktop](https://docs.podflare.ai/integrations/claude-desktop.md): Podflare as an MCP server for Claude Desktop.
- [OpenAI Agents SDK](https://docs.podflare.ai/integrations/openai-agents.md): Drop-in code-execution tool for OpenAI Agents SDK.
- [OpenClaw + ClawHub](https://docs.podflare.ai/integrations/openclaw.md): Install Podflare as a one-line OpenClaw skill from ClawHub. Gives any OpenClaw-backed agent (Claude Code, Cursor, Codex, Cline, or OpenClaw itself) a hardware-isolated sandbox.
- [Vercel AI SDK](https://docs.podflare.ai/integrations/vercel-ai-sdk.md): Code-execution tool for `tool({...})` + `generateText` / `streamText`.
- [Introduction](https://docs.podflare.ai/introduction.md): Podflare is a cloud AI-native sandbox. One sandbox = one full Linux VM with network, filesystem, and a live Python REPL. Hardware-isolated, 80 ms fork, ~190 ms round-trip.
- [Quickstart](https://docs.podflare.ai/quickstart.md): Install an SDK, create a sandbox, run code.
- [MCP server](https://docs.podflare.ai/sdks/mcp.md): Plug Podflare into any MCP-capable client — Claude Desktop, Cursor, Cline, Zed, Windsurf.
- [Python SDK](https://docs.podflare.ai/sdks/python.md): from podflare import Sandbox
- [TypeScript SDK](https://docs.podflare.ai/sdks/typescript.md): import { Sandbox } from 'podflare'