Agent environment / codex

Run Codex in an isolated cloud sandbox.

One API call gives OpenAI's coding agent a real Linux machine — repo checked out, tools installed, network locked down, and nothing left behind when the session ends.

Start building Quickstart

Prefer a full walkthrough? Run Codex in an isolated sandbox →

TEMPLATE

codex

AUTH

OPENAI_API_KEY

ISOLATION

gVisor

EGRESS

default-deny

OVRIN / SANDBOX 9A21 EXITED 0

import ovrin client = ovrin.Client() sandbox = client.sandboxes.create( template="codex", env={"OPENAI_API_KEY": key}, ) result = sandbox.run( 'codex "find and fix the failing test"' ) print(result.stdout)

✓isolated with gVisor19 ms

✓sandbox ready · /workspace mounted151 ms

✓Codex started2.0 s

✓task completed27.4 s

Why a sandbox

Codex is happiest with a whole machine.

Codex is built to read a repository, run the test suite and patch code in a loop. Left alone on a shared host it will happily install packages, spawn subprocesses and rewrite files no one asked it to touch.

01

Untrusted repos by default

Point Codex at a PR from an external contributor without giving it a lane onto your infrastructure.

02

Iterate-and-test loops stay contained

Codex runs the suite, edits, reruns — dozens of times per task. Every process lives in one disposable sandbox.

03

Network on a leash

Default-deny egress. Allowlist api.openai.com, your registry and your git host. Nothing else resolves.

What's in the template