Is idapt Code a separate chat application?

No. It is a curated coding workflow in normal idapt Chat, using the same models, files, computers, tools, history, and workspace permissions.

Does it require GitHub or GitLab integration?

No. The agent uses the Git tools already installed on the computer. If authentication is needed, put the required token in workspace Secrets and refer to it by name.

Does it require VS Code Server?

No. The agent works through the computer shell and filesystem. VS Code Server is optional when you want a browser editor alongside the agent.

Can it use a local model?

Yes, on a paired computer with supported local inference configured. You can also use available hosted models or your workspace provider connections.

Code

Code on your computer or ours.

Use any model in normal idapt Chat to inspect, edit, test, and deliver code.

Start coding Read the quickstart

Where do you want it to run?

Four decisions, four pages. Everything else is the same Chat, the same agent, and the same computer tools.

AI coding on a computer you control

Pair a computer with idapt, start the idapt Code skill in Chat, and let the agent work through the existing shell and filesystem tools.

A fresh cloud computer for the task

Create an idapt cloud computer, clone or upload a project, and run the same idapt Code workflow without preparing local hardware.

Give a coding agent only the access the task needs

Understand the security boundary for paired computers, cloud computers, workspace Secrets, Git credentials, commands, and public preview tunnels.

Use the model that fits each coding task

Run idapt Code with available hosted models, workspace provider connections, or compatible local models while keeping one Chat and computer workflow.

Already using a coding agent?

Both comparisons state what the other product does better, with a review date.

idapt Code vs Claude Code

Where a coding-first terminal harness fits better, and where model choice plus your own machine wins.

idapt Code vs OpenAI Codex

Connected CLI, IDE, and GitHub surfaces compared with one Chat that already holds the rest of your work.

What you get

Your computer or the cloud

Pair a machine you control or create an isolated cloud computer.

Normal idapt Chat

Start with the idapt Code skill. No separate coding interface to learn.

Any available model

Choose cloud models, your provider connections, or local models on compatible hardware.

Real development tools

The agent uses the shell, files, Git, tests, package managers, and preview tunnels.

Credentials by name

Store tokens in workspace Secrets and grant them only to the command that needs them.

Parallel work

Delegate bounded tasks to subagents and keep independent work moving in the background.

Who it's for

Solo developers

Fix, test, and ship from a normal chat without moving the repository off your computer.

Teams

Use workspace-scoped models and Secrets while each task runs on the intended machine.

Private environments

Use a paired computer and local inference when code or prompts should stay on hardware you control.

How it works

1
Choose a computer
Pair your own machine or create a cloud computer from the Computers page.
2
Start idapt Code
Open the curated skill in a new Chat and describe the task and project path.
3
Review the work
Follow commands and results in Chat, then ask for the diff, tests, commit, push, or pull request.

Frequently asked

Give it a project and one concrete task.

Pick a paired or cloud computer, name the project, and say what done looks like.

Start coding Read the docs

Read the docs

Pair with