Hey everyone,

Last week, we released a big feature: Background Computer Use. This week we made it so much faster 🚀 

Fetching data from your apps now takes 77% less time, and acting on it takes milliseconds instead of seconds.

But that’s not all.

We’re now supporting your ChatGPT subscription directly in Alter.

You can use your favorite OpenAI models to power Alter, including their latest GPT 5.6 Sol model.

How to connect your ChatGPT subscription

Open Alter’s settings and go to the API Keys tab. Click Add provider and choose ChatGPT. Then signin with your account to allow Alter.

Once connected, click Manage Models to choose the models you want to see in Alter.

To choose one, type “/chatgpt” in the Alter promptbox, in the hub or in the quickhub.

We also updated the tool calls UI and made several improvements. As usual, let us know what you think on Discord or by email.

Cheers,
Olivier & Sam

Version 2.0.0-beta101: Codex Sign-In & GPT 5.6

TL;DR: Sign in with Codex and use the new GPT 5.6 models. Computer use is dramatically faster — screen reads up to 77% quicker, and clicks/typing resolve in milliseconds instead of seconds on heavy apps like Notes and Spotify. Queue messages to steer your agent mid-task, and follow along with a cleaner grouped tool-call UI.

How to connect your ChatGPT account to Alter

Full Changelog

New Features & Enhancements

• Codex sign-in: New Codex provider integration with support for GPT 5.6 latest models

• Faster computer use: Major AppSense optimizations measured on real apps — AX screen reads up to 77% faster (Notes: 4.4s → 1.7s); element resolution for clicks and fills collapsed from 1–4+ seconds to ~1–4 ms on Spotify, Notes, and Reddit (~330–4100× faster). Plus improved app guidance and dynamic app-specific tools.

• Music app tools: Focused Spotify and Apple Music tools, plus a Deezer recipe.

• Queue messages: Send follow-up messages while the agent is still streaming to steer it mid-task.

• Grouped tool calls: Tool calls are now grouped in conversation view with cleaner summaries, attachments, and styling.

• Reasoning effort selector: Choose reasoning effort for supported models (Codex only, Alter next).

• Send shortcut: Configurable keyboard shortcut for sending messages.

• New Models: GPT 5.6 Luna & Terra

Bug Fixes & Stability

• Hub: Respects IME composition for Enter handling; fixed crash when opening QuickHub and text input undo crash.

• Ask user tool: Limited to 5 options per question.

• Audio: Fixed audio recording on input switch and a memory leak

• Search: Faster search index listing.

• Switcher: Ignores pasted switcher command tokens.