One workspace.
Your entire AI team.

Bring Codex, Claude Code, and Kiro together.
Run agents in parallel,
then review their changes in one place.

Download for Windows Explore the workspace

Your accounts. Your models. Your local projects.

THREE PROVIDERS. ONE PROJECT.

Codex

Claude Code

Kiro

Your local workspaceSessions · Tasks · Changes

Choose a provider for each task.

EXPLORE THE WORKSPACE

DESKTOP / WINDOWS

atcode.acme-platform / Team leader

TEAM WORKSPACE

Dashboard implementation plan

main

Youjust now

Build the project dashboard. Make it clean, connect the API, and cover the main flows with tests.

Team leaderClaude Code

I'll split this into three focused tasks. Each worker gets its own branch so the changes stay easy to review.

Build the dashboard layoutCodex · isolated worktree

Connect the project APIClaude Code · isolated worktree

Cover the happy pathKiro · isolated worktree

The plan is ready. You decide when to start.

Give your team a direction…

CLI default

mainProject indexedCodex 24%Claude 42%Kiro 480 / 2k

Interactive workflow preview · sample project and usageChoose a tab to explore.

WORKS WITH YOUR EXISTING
PROVIDER ACCOUNTS.

Codex

Claude Code

Kiro

CHOOSE A MODEL
FOR EACH TASK.

01 / THE WORKSPACE

Your tools,
in the same project.

Manage provider sessions, parallel tasks, and code review
without switching between terminal windows.

Your accounts and models

Keep using the providers you know. Choose a model for each session, leader, or worker, and switch when the task calls for it.

CODEX · CLAUDE CODE · KIRO

Parallel agents, scoped tasks

Give your team a goal, break it into scoped tasks, and let agents work in isolated Git worktrees. Review the changes before they land.

PLAN → DELEGATE → REVIEW

Terminals beside your code

Run your tools in integrated terminals, keep project context close, and move between conversations and changes without losing your place.

LOCAL FILES · NATIVE TERMINALS

02 / PARALLEL AGENTS

Separate branches.
One review queue.

A team leader keeps the plan in view. Workers tackle focused tasks on separate branches. You stay in control of what gets merged.

Isolated worktrees for parallel work
Project memory and shared task context
Diff review and approval controls

Build your first team

PLAN → WORK → REVIEW

Team leaderPlans the work. Keeps the context.

PLANNING

UI & componentsagent/interfaceScoped task

API & integrationagent/apiScoped task

Tests & checksagent/testsScoped task

Review before mergingApproval required

An example team workflow. Your project stays the source of truth.

YOUR USAGE, IN VIEWEXAMPLE BALANCES

Codex24%

Session limit

Claude Code42%

Weekly limit

Kiro480 / 2,000

Monthly credits used

Illustrative balances. Availability and limits depend on your provider.

03 / PROVIDER USAGE

See your usage.
Plan the next task.

Keep an eye on provider usage without leaving your project. See supported session limits, weekly allocations, and Kiro credit balances together.

Refresh when you need to. Hide what you don't. Your existing provider accounts stay in charge of billing and model access.

A few things to know

04 / GETTING STARTED

From local folder
to your first agent.

Read the setup guide

Bring your project

Open a local folder. Your sessions, tasks, and project memory stay with it.

Choose your team

Use an installed provider, select a model, and give your agents a clear goal.

Review the changes

Follow the work, inspect the diff, and approve the changes you want to keep.

05 / QUESTIONS & ANSWERS

A few things
before you build.

Accounts, code access, and installation.

Getting started

Can I use the AI accounts I already have?

Yes. Atcode works with your installed Codex, Claude Code, and Kiro CLIs and their existing sign-ins. Install and authenticate at least one provider before starting an agent. Model access, usage limits, and charges are determined by your provider and plan.

Do I need all three providers?

No. Start with the provider you already use. You can choose different providers and models for sessions, team leaders, and workers, or leave the model set to CLI default.

Where does my code go?

Projects, saved sessions, tasks, and settings stay on your machine. When you run an agent, the selected provider's CLI processes the relevant prompts and project context according to its own policies. Local storage does not mean that AI requests run offline.

How do parallel agents avoid stepping on each other?

Workers use separate Git worktrees with scoped tasks. You can review their committed changes and approve merges. Isolated workers require a Git repository with at least one commit. Optional project YOLO mode enables automatic worker starts and reviewed merges.

Which operating systems can I download it for?

The current release is available for Windows x64 as a setup executable or MSI installer. macOS and Linux installers are not currently published. Future Windows updates can be downloaded and installed from inside Atcode.

Does Atcode come with AI credits?

Provider usage is billed to your own account. Atcode does not include, pool, or transfer credits between providers. Usage meters help you see the limits and credit balances reported by your connected accounts.

ATCODE FOR WINDOWS

Start with the project
on your machine.

Install Atcode, connect a provider, and open a local folder.

Download for Windows Need the MSI installer?

VERSION 0.1.1·WINDOWS x64·BRING YOUR OWN AI ACCOUNTS

One workspace.Your entire AI team.

Explore the Atcode workspace