QuickTane documentation

QuickTane gives your application a safe place to run code it cannot trust — LLM tool calls, user-submitted snippets, data-analysis scripts. Every execution runs in a fresh, hardware-isolated gVisor sandbox on our own infrastructure, then disappears. No servers to manage, no containers to babysit, no escape surface to worry about.

Two primitives

Everything in QuickTane is built on two ways to run code:

• Runs (one-shot) — submit code, get the result. Ideal for a single LLM tool call or a stateless job. Read more →
• Interactive sessions — a live sandbox that stays up across many exec and file calls, so process and filesystem state persist between steps. Ideal for multi-step agents and REPL-style work. Read more →

Your first run

Authenticate with an API key from your dashboard and POST some code:

curl https://api.quicktane.com/v1/run \
  -H "Authorization: Bearer sk_live_your_key" \
  -H "Content-Type: application/json" \
  -d '{"language": "python", "code": "print(1 + 1)"}'

See the Quickstart for the full request/response cycle, or jump to an SDK:

from quicktane import QuickTane

qt = QuickTane("sk_live_your_key")
run = qt.run_and_wait("print('hello from the sandbox')", language="python")
print(run.output)   # "hello from the sandbox\n"

Where to go next

• Quickstart — make your first call end to end.
• Authentication — API keys and how requests are authorized.
• Interactive sessions — stateful, multi-step execution.
• SDKs — Python and Node/TypeScript clients.
• API reference — every endpoint, generated from the live API.