Product1 publisher3 min readPublished
Sign in with ChatGPT brings a user's AI allowance into 16 partner apps
OpenAI's new Sign in with ChatGPT lets people carry their identity and existing AI allowance into third-party apps, starting with 16 launch partners. For product teams, new users arrive signed in with AI budget to spend, and OpenAI holds both.
The Product Desk · Product desk

What happened
- ChatGPT will suggest an app partway through a conversation when it recognizes the app could help finish a task, and the user can connect and use it without leaving ChatGPT.
- OpenAI expanded ChatGPT's plugin architecture to support extensions, interactive panels where users work in a developer's tool while they keep chatting.
- Plugin review, which TechCrunch likens to Apple's App Review, now lets developers track status, see required fixes, request a human reviewer and update tools without resubmitting.
- Eligible enterprise customers can apply part of their OpenAI commitment toward approved software in a new marketplace that opened with more than 30 partners.
- Each of OpenAI's new Dots agents runs on its own cloud computer and browser and can use the apps a user has connected.
Compiled by The Product DeskSomething wrong?How this is made
Why it matters
- decision Product teams now have to decide whether their core task can be finished inside a chat panel, because that panel is the unit ChatGPT suggests and runs.
- exposure Apps built on the ChatGPT sign-in are exposed to OpenAI's plan decisions: any change to a user's ChatGPT allowance changes what that user can do inside the partner's product.
- cost Enterprise vendors left off OpenAI's approved list will sell against partners that customers can pay with money they have already committed to OpenAI.
- constraint Teams shipping into ChatGPT add OpenAI's review queue to the gates between a release and its users, so OpenAI's reviewers set part of the release schedule.
In the Dev Day demo [1], a user in ChatGPT sees OpenAI's own meeting app list the upcoming meetings on their calendar, lets the AI take notes, and gets a summary of follow-up items when the meeting ends [8]. None of it left the chat [8]. TechCrunch concluded from the day's announcements that ChatGPT is becoming the place where software gets discovered and used, by people and by agents [2].
OpenAI gives ChatGPT's audience as 1.2 billion weekly users [5]. So far, what users have been shown doing is that meeting flow, plus examples of revising a project in Figma and Adobe apps without leaving the conversation [8][9]. TechCrunch's report does not include figures on how often people accept a suggested app or come back to it after the first task. The same report notes that this is not OpenAI's first attempt at putting apps inside the chatbot [16]. It adds that ChatGPT has yet to make full use of its potential as a place to find apps [17].
TechCrunch says the allowance is what gives users a reason to take the suggestion [18]. The launch partners include Cognition's Devin, Notion, Vercel, T3, OpenClaw and Dactyl [4]. The sign-in also reaches beyond the first user. People can share apps with colleagues through the lightweight sites ChatGPT now builds, and each coworker signs in with their own credentials and permissions [10].
Enterprise customers can put committed OpenAI money only toward software OpenAI has approved [13]. Adobe and Figma are in both programs: they appear as design apps shown working inside the chat and as marketplace partners [1].
Agents move users the same way. TechCrunch argues that Dots nudge users to stop thinking about which app to use and to rely on their agent's suggestions [15]. That leaves a product team needing ChatGPT to suggest its app before any user sees it [6].
I think most teams should build one extension that finishes one task inside the panel [7], and keep their own sign-in available next to ChatGPT's. The more of a product that runs on ChatGPT's sign-in and allowance, the more of ChatGPT's audience it can reach, and the less of the customer relationship the team holds [3].
Two questions sort the decision. Does the user's task finish inside a chat panel? Does the buyer already spend through OpenAI?
- Finishes in a panel, buyer spends through OpenAI: build the extension and apply to the marketplace. - Finishes in a panel, buyer pays you directly: build the extension as a way to be found, and keep accounts and billing on your side. - Needs the full product, buyer spends through OpenAI: take a marketplace listing and leave the product where it is. - Needs the full product, buyer pays you directly: wait until launch partners can show how many suggested users came back.
What to watch
- Whether OpenAI explains how ChatGPT picks which app to suggest mid-conversation, and whether that placement can be bought.
- Who covers a partner app's AI usage once a user's ChatGPT allowance runs out, and whether developers pay OpenAI anything for it.
- Return-rate figures from the 16 Sign in with ChatGPT launch partners, and how quickly OpenAI adds more.