OpenAI tells developers ChatGPT is ready to be their gatekeeper

1 month ago 5

OpenAI on Monday pitched its coding tools to software developers in the hope of generating the usage and revenue necessary to recoup the vast sums it spends to create and run its AI services.

CEO Sam Altman presided over OpenAI DevDay 2025 at San Francisco's Fort Mason Center, the third such event since the company began chewing through massive amounts of costly cloud computing to resell scraped data as service.

"Today, 4 million developers have built with OpenAI," said Altman in a streamed presentation. "More than 800 million people use ChatGPT every week, and we process over six billion tokens per minute on the API, thanks to all of you. AI has gone from something people play with to something people build with every day."

That hasn't been enough to turn a profit. According to The Information, OpenAI generated $4.3 billion in revenue in the first half of 2025, 16 percent more than all of 2024, and reported a net loss of $13.5 billion, up from $3.1 billion during the same period a year ago.

OpenAI did not respond to a request to confirm those figures.

But Altman insists that there's never been a better time to build software applications with OpenAI, and he provided some details on how the AI giant aims to help developers do that.

Altman and underlings went on to describe:

Apps SDK, a way to integrate apps into ChatGPT
AgentKit, a set of tools for creating AI-based agents, including Agent Builder, Connector Registry, and ChatKit
The addition of more Evals for testing and measuring model behavior
The general availability of Codex
The debut of the GPT-5 Pro API
And a smaller voice model called gpt-realtime-mini.

Here's a rundown of the most important and interesting bits.

Apps SDK allows ChatGPT to send users to your program

Apps SDK provides a way to discover and invoke apps through the ChatGPT interface.

"When someone's using ChatGPT, you'll be able to find an app by asking for it by name," Altman explained. "For example, you could sketch out a product flow for ChatGPT and then say 'Figma, turn the sketch into a workable diagram.' The Figma app will take over, respond, and complete the action."

(In a sign of how desperate investors are to get in on the AI boom, Figma's stock went up 16 percent after Altman talked about them in his presentation.)

Functionally, the idea is similar to an Android Intent – a way to pass data from one mobile app to another. But OpenAI aims to generalize this sort of cross-app communication so it works on a broader level, through natural language. And it is doing so in a way that will have economic implications for app developers.

To participate, developers have to set up metadata that informs ChatGPT about their app's functions, create an MCP server to handle communication with ChatGPT, and connect the MCP server to ChatGPT. They have a substantial incentive to do so given the size of the ChatGPT audience – integration is a prerequisite for having ChatGPT surface apps. However, app makers who agree to entrust distribution to a third party may find themselves at the mercy of that entity, as Apple's App Store and Google Play have demonstrated.

Developers comfortable with that tradeoff will be able to register their apps so that ChatGPT can route queries to those apps when the topic and metadata suggest a match.

Alexi Christakis, a member of OpenAI's technical staff, explained how ChatGPT app discovery works when the user hasn't tried the app before; it requires user consent to connect the app to your ChatGPT conversation.

AgentKit helps create workflows

AgentKit provides a way to configure workflows to be carried out by agents (models running tools in a loop). The Agent Builder offers visual flowcharting to connect models with data and tools. The Connector Registry provides an administrative panel governing how ChatGPT and the API get linked to other services, such as pre-built connectors for Dropbox, Google Drive, Sharepoint, and Microsoft Teams, and third-party MCP servers.

"AgentKit is a complete set of building blocks available in the OpenAI platform designed to help you take agents from prototype to production," said Altman. "It is everything you need to build, deploy, and optimize Agentic workflows with way less friction."

Altman described how grocery chain Albertson's built an agent using AgentKit. Faced with a scenario in which ice cream sales had dropped unexpectedly by 32 percent, the grocery chain would typically have undertaken a long process of reporting and spreadsheet data analysis, he said.

"Now, an associate can just ask the agent what's going on," Altman said. "The agent will look at the full context of everything it can discover – seasonality, historical trends, external factors – and it will give a recommendation."

He didn't address whether the resulting recommendation would reverse the sales slump or how it might compare to slower human data analysis.

The utility of AI intervention was also left unaddressed in a demonstration of Agent Builder by OpenAI technical staff member Christina Huang. She created an agent to recommend sessions at OpenAI's DevDay in less than eight minutes. Whether this low-code experience produced something of value depends on whether this is the sort of task one prefers to delegate to a hastily-assembled web page widget or whether it's the sort of challenge one might manage without aid.

Altman concluded, "We want OpenAI to be a great platform for this new era of building. We think things are going to get pretty incredible pretty soon."

For the sake of OpenAI's investors, please hurry. ®

Read Entire Article