Feature Request: Replace Codex Pet with a Codex Task Board

I’d like to suggest a change to the Codex experience.

As Codex becomes increasingly capable of working on multiple coding tasks and projects, I think there is a growing need for a task management layer on top of the existing execution experience.

The current Codex Pet experience is fun, but personally, I would get much more practical value from a Codex Task Board / Agent Control Center that lets me manage, monitor, and organize all of my Codex tasks in one place.

The problem

As I use Codex across more projects, I often have many tasks running or waiting at the same time.

For example, I might have Codex working on:

  • Implementing authentication

  • Fixing a payment webhook

  • Refactoring an API client

  • Adding unit tests

  • Updating documentation

  • Investigating a CI failure

  • Reviewing an existing implementation

  • Preparing a pull request

When there are only one or two tasks, the current workflow is fine.

But once there are many tasks running in parallel, task management becomes the bottleneck.

I need to be able to quickly answer questions like:

  • How many Codex tasks are currently running?

  • Which tasks are waiting for me?

  • Which tasks have completed?

  • Which tasks failed?

  • Which tasks are blocked?

  • Which repository/project does each task belong to?

  • Which tasks have created a branch or PR?

  • Which tasks need my attention right now?

  • What is each Codex agent currently doing?

Currently, it is difficult to get this information at a glance.

Proposed feature: Codex Task Board

I’d love to see a dedicated Codex Task Board or Agent Control Center.

A Kanban-style interface could work very well:

β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚   BACKLOG    β”‚    RUNNING   β”‚   NEEDS ME   β”‚    REVIEW    β”‚  COMPLETED   β”‚
β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€
β”‚ Update docs  β”‚ Auth API     β”‚ DB decision  β”‚ PR #123      β”‚ Webhook      β”‚
β”‚ Unit tests   β”‚ DB refactor  β”‚ Permission   β”‚ PR #124      β”‚ API client   β”‚
β”‚ Refactor     β”‚ CI fixes     β”‚ Question     β”‚              β”‚ Tests        β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

Each Codex task could be represented as a card.

What could a task card contain?

For example:

β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚ πŸ”΅ Implement OAuth authentication β”‚
β”‚                                    β”‚
β”‚ Repository                         β”‚
β”‚ my-company/backend                 β”‚
β”‚                                    β”‚
β”‚ Status: Running                    β”‚
β”‚ Runtime: 18 minutes                β”‚
β”‚                                    β”‚
β”‚ Branch                             β”‚
β”‚ codex/implement-oauth              β”‚
β”‚                                    β”‚
β”‚ Last activity                      β”‚
β”‚ Implementing OAuth callback...     β”‚
β”‚                                    β”‚
β”‚ [Open Task] [Stop] [Open Branch]   β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

Useful information could include:

  • Task name / original prompt

  • Repository / project

  • Current status

  • Branch or worktree

  • Start time

  • Elapsed time

  • Last activity

  • Current agent activity

  • Relevant commit

  • Pull request

  • Test status

  • Short summary of the current task

  • Actions such as Resume / Review / Stop / Archive

The goal would not simply be to create another Kanban UI.

The goal would be to provide a single control center for all Codex work.

A very important status: β€œNeeds Me”

One status that I think would be particularly valuable is:

Needs Me

Sometimes an agent cannot continue without user input.

For example:

  • Codex needs me to choose between two implementation approaches

  • Codex needs permission

  • Codex needs clarification

  • Codex encounters a conflict

  • Codex needs a configuration value

  • Codex has finished implementation and is waiting for review

Instead of discovering these situations one by one, I could simply open Codex and see:

Codex

🟒 8 Running
❓ 2 Need Your Input
πŸ‘€ 3 Need Review
❌ 1 Failed
βœ“ 14 Completed

That would immediately tell me where my attention is required.

Repository and project filtering

As the number of tasks grows, filtering would also become important.

For example:

Repository
β”œβ”€β”€ frontend
β”œβ”€β”€ backend
β”œβ”€β”€ mobile
└── infrastructure

Status
β”œβ”€β”€ Running
β”œβ”€β”€ Needs Me
β”œβ”€β”€ Review
β”œβ”€β”€ Failed
└── Completed

I could then quickly answer:

β€œShow me everything currently running in the backend repository.”

Or:

β€œShow me every Codex task that needs my attention.”

GitHub / Pull Request integration

I think this could become especially powerful if a Codex task were connected to the entire development lifecycle:

Codex Task
    ↓
Repository
    ↓
Branch
    ↓
Commits
    ↓
Pull Request
    ↓
Code Review
    ↓
Merged

For example:

Implement OAuth

Status: Review

Repository:
my-project/backend

Branch:
codex/oauth

Commits:
3

Pull Request:
#184

Tests:
βœ“ 127 passed

[Open PR]
[Open Task]
[Continue with Codex]

This would make Codex feel less like a collection of independent coding sessions and more like an actual AI software engineering workspace.

Batch management

If I have many tasks, it would also be useful to select multiple tasks and perform actions such as:

  • Stop

  • Resume

  • Archive

  • Retry

  • Open

  • Move to another status

For example:

Select 5 tasks

[Resume] [Stop] [Archive]

This would become increasingly valuable as Codex supports more parallel agents.

Why I think this matters

I think the direction of Codex is increasingly agent-oriented.

The more capable Codex becomes at working independently and in parallel, the less useful it is to think of Codex as just a single coding session.

The new problem becomes:

How do I manage all the work that Codex is doing for me?

In other words:

Codex needs a task management layer, not just an execution interface.

A Task Board could provide that layer.

It could make Codex feel more like:

An AI software engineering workspace where I can manage a queue of work, monitor multiple agents, review their results, and intervene when necessary.

About Codex Pet

I personally would prefer the space currently used by Codex Pet to be used for this kind of functionality.

However, I don’t necessarily mean that Codex Pet must be completely removed.

If the team wants to keep the Pet experience, another possibility would be to make it optional or provide a switch between:

Codex Pet ↔ Codex Task Board

The important part for me is having a proper place to manage multiple Codex tasks.

As Codex becomes more capable of handling parallel work, I think this kind of task-management experience could become an important part of the product.

I’d love to see a Codex Task Board / Agent Control Center considered for the future of Codex.

Why β€˜replace’ the pet with a completely different feature, what did the pet do to deserve this? Just add the feature on.

My dog is displeased.

the dog is cute.

but codex pet is useless

Have you looked at Codexometer: track your quota, command sessions & benchmark models!? Allows you to monitor multiple sessions in one dashboard and to visualise and respond immediately to any awaiting approvals or complete turns.