Runtime
The SDKs, CLI, MCP server and agent integrations for Runtime: sandboxes for AI agents. Each sandbox is a Firecracker microVM with its own Linux kernel, started from one call.
npm install withruntime # TypeScript SDK, the `runtime` CLI and the MCP server
pip install withruntime # Python SDK, sync and async
import { Sandbox } from "withruntime";
await using sbx = await Sandbox.create();
const result = await sbx.exec("python3 -c 'print(6 * 7)'");
console.log(result.exitCode, result.stdout);
from withruntime import Sandbox
with Sandbox.create() as sbx:
result = sbx.exec("python3 -c 'print(6 * 7)'")
print(result.exit_code, result.stdout)
The client reads RUNTIME_API_KEY, or the connection npx withruntime login
saved on this machine. A new account starts on a free trial with no card. The
quickstart takes it from there.
What is here
Documentation
- Start
- TypeScript and Python
- CLI and MCP
- Frameworks, and moving from E2B or Daytona
- Pricing
How Runtime compares
Pages that answer what people ask when choosing a sandbox, each with the providers' published rates and the date they were read:
- Best sandbox for AI agents
- Alternatives to E2B, Daytona, Modal and Vercel Sandbox
- E2B vs Daytona, and every comparison
- The agent sandbox price index: every provider's rates, also as CSV and JSON
- Sandbox cost calculator and Runtime's pricing
Tests
bun test in packages/cloud-sdk and python3 -m unittest discover -s tests
in sdks/python run here as they are. The suites that start Runtime's API
server in-process, including all of the Go, Java and Ruby tests, run in
Runtime's main repository before each release.
Contributing
Each release copies the SDKs here from Runtime's main repository, so each commit here is one copy, named after the versions in its source; a copy made before those versions reach npm and PyPI says so. Issues are welcome. A pull request is reviewed here and carried into the next release with its author credited.
License
Apache-2.0. See LICENSE.