Unlock Codex’sfull potential
Connect your models, discover features, and manage tasks and costs with Codex Pass.
Keep the client you know and make Codex part of your work.
Make Codex work for you
Use your official account alongside your API, choose models, adjust Codex settings and inspect the native environment and Skills.
01Original experienceKeep the ChatGPT you know
Bring in new models and keep the client features you use every day.
Keep the ChatGPT you know
Bring in new models and keep the client features you use every day.
Keep using voice,
images and more.
Connecting a model should not mean learning another chat app. Codex Pass extends the original client from the outside, so everyday work stays in a familiar interface.
Voice and images
Continue using voice and image generation through your official account, in the same client.
Search and remote access
Keep the client’s native web search and remote control, with the interactions you already know.
Two sources of capability
Native features come from your official account. Custom models use your configured API. Each follows its service permissions.
02ConfigurationMake Codex fit the way you work
Bring scattered features and settings into one place to explore and adjust.
Make Codex fit the way you work
Bring scattered features and settings into one place to explore and adjust.
Discover what is possible.
Choose what you need.
Codex has capabilities that are easy to miss. Codex Pass brings hundreds of configurable features together, making discovery part of using the product.
Find features
Explore what Codex can do beyond its default setup.
Configure for your tasks
Choose features and settings around the work you need to do.
Build your setup
Bring feature choices, models and working preferences together in your own Codex configuration.
03Third-party modelsStay in control of your model choices
Use your API and choose around cost, capability and the task at hand.
Stay in control of your model choices
Use your API and choose around cost, capability and the task at hand.
Your choice of models.
Inside a familiar client.
Already have a model provider you like? Connect its service URL, API key and models, then keep working in Codex with the model your task needs.
Bring your provider
Connect an API from a model provider or AI gateway and use the resources you already have.
Choose for the task
Routine work and complex tasks can use different models. Weigh the provider’s pricing and model capabilities.
Keep native features
Custom APIs handle requests to your selected models. Native features such as voice and images still follow your official account’s capabilities.
04Delegation and managed tasksGive each model a role in the work
Move from a single model to collaboration with distinct responsibilities.
Give each model a role in the work
Move from a single model to collaboration with distinct responsibilities.
Plan, execute and review
with different models.
A task often involves several kinds of work. Assign models to different roles, and use managed tasks for work that needs to keep moving.
Assign roles
Use different models for planning, execution and review within a task.
Manage ongoing work
Use automatic task management for work that needs sustained progress.
Learn from usage
Review logs and costs to understand model usage and adjust future assignments.
05Logs and costsSee the calls and costs behind a session
Follow usage from individual requests to the project.
See the calls and costs behind a session
Follow usage from individual requests to the project.
Understand what
a project consumed.
After a task finishes, you may still want to know which calls ran and what they used. Codex Pass makes logs, sessions and project costs available to inspect and learn from.
Request logs
Inspect call details when investigating failed or unexpected requests.
Session history
Revisit conversations and task context to understand how work unfolded.
Project costs
Review costs by project to understand how different work consumes resources.
06Codex DoctorStart troubleshooting with your environment
Check configuration, connections and runtime in one diagnostic flow.
Start troubleshooting with your environment
Check configuration, connections and runtime in one diagnostic flow.
Give Doctor
a first look.
When a model will not connect or a tool stops working, the cause can be hard to locate. Doctor checks your AI development environment to help narrow the investigation.
Check the environment
Inspect the local environment your AI tools depend on for potential problems.
Check configuration and connections
Use configuration and connection checks to find clues behind failed calls.
Follow the findings
Use the diagnostic results to focus further troubleshooting and reduce trial and error.
Native features remain subject to your official account’s permissions and supported capabilities.
Deploy Codex
for your team
Everything in the personal edition, plus resource management and usage controls for team collaboration. Administrators configure providers and assign models and usage quotas to members.
Open Console- 01
AI gateway access
Connect model providers and AI gateways centrally. Members access models with a connection code, without receiving provider API keys.
- 02
Budgets & model access
Assign available models and usage quotas per member to match their work.
- 03
Team memory
Save project conventions and lessons so colleagues’ Codex can draw on shared knowledge.
- 04
Task progress
View running tasks, responses and interruptions to understand progress and issues that need attention.
- 05
Sharing
Share conversations with invited colleagues so they can continue the discussion with context.
- 06
Remote collaboration
Away from your desk, connect to the original device to inspect and continue ongoing tasks.
Use models to inform
selection and judgment
Which model fits a task? Does a piece of content carry potential risk? Jev brings model-based judgment to model planning and harmful-content assessment.
Bring agents into
everyday work
Advanced AI capabilities should be within reach of more people, in the work they do every day.
Codex provides a strong foundation for collaboration between people and agents. We extend that experience without modifying the original client, bringing model access, task collaboration and usage controls into daily work so individuals and teams can use AI on their own terms.
Progress comes from sustained practice in real settings. We turn what we learn into reusable tools and invite people across industries to help shape them, making Codex a capability that more professionals and organizations can put to work.
Problems from daily use shape what we build next.
No invite code required. Get started today.
Team setup, model and quota management, and connection security.
Register and connect your team
Team onboarding
From registration to team access
Register your team in the console. Administrators configure providers in the console and assign models and usage quotas to members.
Recommended preparation: A Business / Team plan, plus accounts and API credentials for the model providers you intend to connect.
01 Team owner
Register your team
Open the console, register and choose an available plan for your team.
02 Administrator
Configure providers
Connect model providers or AI gateways in the console. Manage service URLs, API keys and available models centrally.
03 Administrator
Assign models and quotas
Set which models each member can call and assign usage quotas based on their work.
04 Member
Connect and start using
Members connect with a connection code and use their assigned models within their quotas. Administrators do not need to distribute provider API keys to members.
Models and protocols
Models & routing
Team models in the Codex model picker
Team models are available in the existing Codex model picker, allowing members to retain their current workflow.
- Model and pricing information. Models and their prices are displayed directly in the Codex model picker.
- Local protocol conversion. Supports Responses, Chat Completions and Anthropic formats, with local protocol conversion.
- Automatic fallback. If a gateway doesn't support Responses, requests switch to Chat Completions.
- Individual use. With Custom API, enter a URL, key and model, and test it before saving.
Team console
Team console · Available now
Centralized team administration
Use console.cp.dev for workspace creation, channel configuration, connection codes, member allowances and usage monitoring.
-
App connectionsIssue connection codes with project binding and revocation controls.
-
Plan & usageManage credits, member allowances and trial status centrally.
-
Channels & modelsConfigure routing rules and model access permissions for members.
-
Budgets & alertsReceive alerts through Feishu, WeCom, DingTalk, webhooks or email as projects approach their budget limits.
-
DLPRedact or block email addresses, phone numbers, identity and card numbers, and keys before requests are sent.
-
Push projects and threadsDistribute projects and conversations to member devices, with direct access through Open now.
-
Project Memory, backups & handoversTransfer project knowledge and work in progress during personnel handovers.
Connection and credential protection
Security
Connection and credential protection
Local access restrictions, device authentication, end-to-end encryption and signature verification protect connections and credentials.
- Local onlyThe local proxy listens only on 127.0.0.1 or ::1 and uses a per-process token. It is not exposed to the local network or public internet.
- Signed by the deviceEach device uses its own Ed25519 key to sign sync and inference requests for device authentication.
- One-time connection codesConnection codes have a short validity period. Decryption keys remain on the local device, and failed setup triggers a complete rollback.
- End-to-end encrypted remoteUses an X25519 handshake and AES-256-GCM encryption. Keys are stored in the URL fragment after # and are not sent to the relay server.
- Invite-only sharesShare pages prohibit search engine indexing. The sharing service does not receive connection codes, model keys or ChatGPT sign-in credentials.
- Signed buildsThe macOS app is signed and notarized, the Windows build carries an Authenticode signature, and updates are verified before they install.
FAQ
A few things you might ask.
Can I start without an invitation?
Yes. Download Codex Pass and register without an invitation code. Features that use your own resources remain free. Visit the console for hosted services, team capabilities and available plans.
What are the prerequisites?
Members need the official ChatGPT desktop app with Codex. Teams should register in the console first. We recommend preparing a Business / Team plan and API credentials for a provider or AI gateway. Once enabled, administrators configure providers, model access and usage quotas centrally, without distributing API keys to members.
Does the existing Codex workflow change?
No. Codex retains its existing interface and workflow. Team models are available directly in its model picker.
How are requests routed and processed?
In team mode, Codex sends requests to Codex Pass on your computer, which signs them with the device key and forwards them to console.cp.dev. The console checks the signature, deducts the allowance and applies DLP rules, then sends them on to the AI gateway or API key you connected. With Custom API, requests go straight to the address you entered.
Which systems are supported?
macOS (Apple silicon and Intel), Windows x64, and Linux x64 and ARM64 (AppImage, DEB, RPM). The app comes in English, Simplified Chinese, Traditional Chinese and Japanese.
Is individual use supported?
Yes. The personal edition preserves the native client experience and adds configuration, model access, delegation, managed tasks, logs, cost visibility and Doctor diagnostics.
CODEX PASS
Download. Start your own work.
Available for macOS, Windows and Linux. No invite code required. Start free and try Personal for 14 days.