Working examples for running code and AI agents in isolated, disposable Linux microVMs with Tenki Sandbox.
Use this cookbook to:
- run untrusted or agent-generated code without exposing your host;
- give popular agent frameworks a secure code-execution tool;
- expose sandboxed services through public preview URLs;
- preserve an environment with snapshots; or
- migrate an existing E2B, Modal, or Daytona integration to Tenki.
New to Tenki? Start with Run code in a sandbox. It covers the core create, execute, and dispose lifecycle in about 15 lines of JavaScript.
You need:
- a Tenki account;
- a Tenki API key and your project and workspace IDs; and
- Node.js 20 or later.
Clone the cookbook and run the introductory example:
git clone https://github.com/LuxorLabs/cookbook.git
cd cookbook/examples/run-code-in-a-sandbox
npm install
export TENKI_AUTH_TOKEN="your-api-key"
export TENKI_PROJECT_ID="your-project-id"
export TENKI_WORKSPACE_ID="your-workspace-id"
node run.mjsThe script creates a sandbox, runs Python inside it, prints 42, and disposes
the sandbox automatically. See the
Sandbox SDK guide for authentication
and client setup details.
| Example | What you will learn |
|---|---|
| Run code in a sandbox | Create a microVM, execute a command, read its output, and dispose it |
| Work with files | Write, read, and list files with the TypeScript SDK |
| Expose a port | Run a web server and create a public preview URL |
| Pause, snapshot, and resume | Preserve a sandbox and resume it with its state intact |
| Run code with Python | Use the official tenki-sandbox Python SDK |
| Example | Framework or use case |
|---|---|
| Vercel AI SDK | Use Tenki as the AI SDK's experimental_sandbox |
| LangChain (JavaScript) | Give a LangChain agent a sandboxed code interpreter |
| LangChain (Python) | Add a Tenki code-execution tool to a Python agent |
| CrewAI | Run agent-generated code in a Tenki sandbox |
| OpenAI Agents SDK | Add sandboxed execution as an agent tool |
| LlamaIndex | Wrap sandboxed execution in a FunctionTool |
| Hugging Face smolagents | Back a CodeAgent with a remote Python executor |
| Vercel Eve | Use Tenki as an Eve agent's sandbox backend |
| Example | Integration |
|---|---|
| MCP server | Give Claude, Cursor, or another MCP client access to Tenki sandbox tools |
| Composio | Add Tenki tools to a Composio agent |
| Covalent | Run each workflow task in its own microVM |
| Guide | What it includes |
|---|---|
| E2B to Tenki | Side-by-side code and an API mapping |
| Modal to Tenki | Side-by-side code and an API mapping |
| Daytona to Tenki | Side-by-side code and an API mapping |
Each example is self-contained. Open its README for prerequisites, complete source code, and run instructions.
Cookbook examples should be useful when you need them, not just when they are
written. Each example includes a verify.mjs script that exercises its
Tenki-facing behavior against the live API. CI runs those checks on every push
and pull request.
To verify an individual example after setting your Tenki credentials:
cd examples/run-code-in-a-sandbox
npm install
node verify.mjsMaintainers and contributors can run every example from the repository root:
npm run verifySee CONTRIBUTING.md for the verification contract and the recommended structure for new examples.
These first-party packages can be installed directly in your own project:
| Integration | Install | What it does |
|---|---|---|
| Composio tools | npm install @tenkicloud/composio-tools |
Provides create, execute, snapshot, and terminate tools for Composio agents |
| Covalent executor | pip install covalent-tenki-plugin |
Runs Covalent workflow tasks in disposable Tenki microVMs |
| GitHub Actions | uses: TenkiCloud/actions/setup-cli@v1 |
Installs the Tenki CLI and builds sandbox templates in CI |
| Go SDK | go get github.com/TenkiCloud/tenki-sdk-go/sandbox |
Provides a Go client for the Tenki Sandbox API |
Contributions are welcome. Read CONTRIBUTING.md for local setup and the example checklist, then choose an idea from ROADMAP.md.