sandbox-agent

Deploy, configure, and integrate Sandbox Agent - a universal API for orchestrating AI coding agents (Claude Code, Codex, OpenCode, Amp) in sandboxed environments. Use when setting up sandbox-agent server locally or in cloud sandboxes (E2B, Daytona, Docker), creating and managing agent sessions via SDK or API, streaming agent events and handling human-in-the-loop interactions, building chat UIs for coding agents, or understanding the universal schema for agent responses.

14stars2forksUpdated 4/30/2026

DevOps & CI/CD Cloud Infrastructure #ai-coding #docker #integration #sandbox #cloud #developer-tools

Security Assessment

Critical Risk(0/100)

Detected risks:

Destructive Operations([references/common-software.md] rm -rf, [references/deploy/docker.md] rm -rf, [references/deploy/boxlite.md] rm -rf)

Persistence([references/computer-use.md] launchD)

Remote Code Execution([SKILL.md] curl -fsSL https://releases.rivet.dev/sandbox-agent/0.4.x/install.sh | sh, [references/deploy/daytona.md] curl -fsSL https://releases.rivet.dev/sandbox-agent/0.4.x/install.sh | sh, [references/quickstart.md] curl -fsSL https://releases.rivet.dev/sandbox-agent/0.4.x/install.sh | sh)

Secret Exposure([SKILL.md] API_KEY=, [SKILL.md] API_KEY =, [references/llm-credentials.md] API_KEY=)

Sensitive File Access([SKILL.md] .env, [references/llm-credentials.md] .env, [references/deploy/daytona.md] .env)

Privilege Escalation([references/cli.md] sudo)

Security Score0/100

About sandbox-agent

The sandbox-agent skill provides a universal API designed to orchestrate AI coding agents such as Claude Code, Codex, OpenCode, and Amp within sandboxed environments. This skill simplifies the deployment, configuration, and integration of these agents, making it easier to set up sandbox-agent servers locally or in cloud sandboxes like E2B, Daytona, and Docker. It also offers tools to create and manage agent sessions, stream agent events, and handle human-in-the-loop interactions. By offering a uniform schema for agent responses, sandbox-agent streamlines working with multiple AI coding agents, allowing users to focus on building applications without the complexity of managing individual agent interactions across different environments.

FAQ

How do I install the sandbox-agent skill?

You can install the sandbox-agent skill using either npx, bunx, or the SDK manager of your choice. For example, use `npx skills add rivet-dev/skills -s sandbox-agent` or `bunx skills add rivet-dev/skills -s sandbox-agent`.

What environment variables do I need to set up?

You need to set up API keys for each coding agent, such as `ANTHROPIC_API_KEY` and `OPENAI_API_KEY`, to connect to their respective providers. These keys can be set in local shells, or dynamically within cloud environments like E2B or Daytona.

Can I test the sandbox-agent without API keys?

Yes, you can use the `mock` agent for SDK and integration testing without requiring provider credentials.

What are the system requirements for running sandbox-agent?

You can run sandbox-agent locally or in cloud sandboxes like E2B, Daytona, and Docker. Specific instructions for each environment are provided in the documentation, including installation through curl, npx, bunx, or Docker.

Is sandbox-agent compatible with multi-tenant setups?

Yes, sandbox-agent supports multi-tenant setups and can handle per-user billing and usage tracking. For more advanced features, you can integrate with LLM gateway options like OpenRouter, LiteLLM, or Portkey.

All Files

47 files

SKILL.md10.0 KB