Sandbox agent workspace

sandbox-agent-workspace

Use SandboxAgent with a Manifest to define a workspace (git repos, local files, URLs) that the agent can inspect and modify. Run with SandboxRunConfig specifying the sandbox client.

When to apply

Long-horizon tasks that need a real filesystem: inspecting repos, editing code, running shell commands, applying patches. Available from SDK version 0.14.0+.

Core snippet

import asyncio
from agents import Runner
from agents.run import RunConfig
from agents.sandbox import Manifest, SandboxAgent, SandboxRunConfig
from agents.sandbox.entries import GitRepo
from agents.sandbox.sandboxes import UnixLocalSandboxClient

agent = SandboxAgent(
    name="Workspace Assistant",
    instructions="Inspect the sandbox workspace before answering.",
    default_manifest=Manifest(
        entries={
            "repo": GitRepo(repo="openai/openai-agents-python", ref="main"),
        }
    ),
)

async def main():
    result = await Runner.run(
        agent,
        "Inspect the repo README and summarize what this project does.",
        run_config=RunConfig(
            sandbox=SandboxRunConfig(client=UnixLocalSandboxClient())
        ),
    )
    print(result.final_output)

asyncio.run(main())

Key notes

• SandboxAgent extends Agent with default_manifest, base_instructions, capabilities, run_as
• Manifest.entries maps names to GitRepo, LocalDir, LocalFile, RemoteFile etc.
• UnixLocalSandboxClient uses the local filesystem; other clients can use Docker or remote environments
• SandboxRunConfig is nested inside RunConfig(sandbox=...)
• The agent has access to ApplyPatchTool, ShellTool, and other workspace-native tools