THE WORKSPACE / WINDOWS

Give your AI
room to work.

VARIANT-1 gives your model a persistent Python environment and tools for files, browsing, and desktop work. Ask for a result, inspect what it produces, and keep going in the same chat.

01 / THE WORKSPACE

From the question
to the working file.

Bring a project, a set of documents, or a task in another application. Your model can use code and tools to work through it, while the chat, files, browser, terminal, and review panes keep the work in view.

Explore example workflows

ONE WORKING ENVIRONMENT

VARIANT-1

Ask → work → review

ChatPythonFilesBrowserDesktopTools

Your model chooses the steps. Python and connected tools give it ways to carry them out.

Development interface showing the chat workspace, files, review, and terminal with synthetic sample data. — Development UI with sample data · September 2026. The release interface may change.

02 / WORKING STATE

Continue from
what you already have.

Clean a dataset, then ask for a different comparison. The model can reuse the data and helper functions from earlier calls while the chat’s Python runtime stays alive.

A restart or lost runtime can clear live state. Optional checkpoints preserve supported values when enabled; saved files and approved memory are stored separately.

See what persists

EXAMPLE / LIVE SESSION

LIVE WORKING STATE

Ready for the follow-up.

recordsavailable

cleaned_recordsavailable

normalise()available

The data and helper are still in Python. The next call can use them directly.

03 / COMPOSABLE METHODS

Build a method.
Refine it as you go.

Your model can turn a repeated step into a Python helper and combine it with other tools. The code stays available to read, change, and use again.

Enable mutation to let the model propose and activate revised tool methods within the chat. These are changes to executable code; the model’s weights stay the same.

Read about tools and mutation

Mutation is off by default. Turning it off stops new changes, but does not remove changes already active in the chat.

METHOD / VARIATION

Compose the pieces.

Read the inputs, process the data, and call the tools needed to produce the output.

Keep a useful method.

Give a repeated operation a name and clear inputs so it can be reused and adjusted.

Check the result.

Open the file, compare the values, or inspect the application. Use the result to decide what needs changing.

04 / MODEL CHOICE

Choose the connection.
Keep the workspace.

Use a supported local model, connect an API key, or sign in through a provider account. Choose the model for your task and keep the same place to work.