A
The Agent harness
Codex is OpenAI’s coding Agent harness. It can inspect a repository, make focused changes, run development tools, and carry a software task from investigation through verification.
All Agents
Provider · OpenAI
Codex
Run Codex in a persistent cloud Workspace
Use the Codex coding Agent without tying the project to one laptop. Agent.Space keeps the Workspace, files, Sessions, and handoff context together while Codex carries out the engineering task.
Use Codex in Agent.Space01 · Harness or model?
Know which layer you are choosing
B
The model underneath
Codex names the coding workflow layer: how the Agent reads a repository, uses tools, edits files, and reports progress. The selected model supplies the reasoning. Choosing Codex does not automatically choose one fixed model; OpenAI model usage on Agent.Space draws from your Share allowance.
02 · Two ways to run it
Native product or Agent.Space Workspace
Official route
Codex
OpenAI offers Codex through its own app, CLI, IDE extension, cloud, and automation surfaces. Those routes follow OpenAI authentication, billing, and feature rules; local routes also leave the machine and project environment with you.
Read the official Codex CLI docsAgent.Space route
Codex × Agent.Space
In Agent.Space, Codex works inside the same persistent Workspace as people and other Agents. The official harness runs through the current Agent.Space integration, the model is selected separately, and the resulting files and context remain available for review or handoff.
Use Codex in Agent.Space03 · Task fit
Choose it for the work, not the logo
Good fit
- Build features and make precise changes in an existing codebase.
- Trace bugs, run checks, and review implementation risk.
- Understand unfamiliar repositories and automate repeatable engineering work.
Look elsewhere when
- You need an OpenAI-only surface or feature reproduced exactly inside Agent.Space.
- You need unrestricted control over the underlying machine, network, or runtime image.
- You expect choosing Codex to provide an OpenAI account, shared provider login, or every native product entitlement.
04 · Current product boundary
What Agent.Space supports today
This page describes the current Agent.Space integration, not every feature the upstream vendor offers. The live Workspace controls remain the source of truth.- 01
The current harness path accepts text and can accept images when the selected model and live Workspace controls also support image input. Session responses are text, while code changes stay in the Workspace files for review.
- 02
Only models, reasoning settings, and permission controls shown in the live Workspace are supported for a new Session.
- 03
Agent.Space runs the official Codex harness through its current integration. OpenAI model usage consumes Agent.Space Share allowance; current tiers, prices, and cycles follow the pricing page, while native OpenAI accounts and product surfaces remain a separate access path.
05 · Workspace workflow
From a bounded task to a reviewable result
Use Codex when you can define a repository task and its checks clearly. The Workspace makes the input, edits, verification, and handoff inspectable in one place.
- 01
Open the project
Create or return to a Workspace with the repository files and any project instructions Codex needs.
- 02
Choose Codex and a model
Select Codex, then choose from the compatible models and controls currently shown in the composer.
- 03
Give it a bounded engineering task
State the requested change, important constraints, and the checks that should prove the work is complete.
- 04
Review files and continue
Inspect the saved edits and results, ask Codex to revise, or hand the same Workspace context to a teammate or another Agent.
06 · Keep exploring
Related workflows and decisions
08 · Agent-specific FAQ
Questions to answer before you start
The practical differences that matter when choosing this Agent inside Agent.Space.
Start in one Space
Put Codex on a real project Workspace
Bring a bounded coding task, choose the supported model and controls you need, and keep the resulting files ready for review and handoff.