OpenAI is now letting users work with third-party apps and complete certain tasks without leaving ChatGPT’s interface.
OpenAI says Codes is now generally available to eligible users
Photo Credit: Reuters
OpenAI DevDay 2025 was hosted by the company on Monday, and multiple announcements for both end-users and developers were made. Among them, the biggest announcement was adding a new capability in ChatGPT that will allow users to work with third-party apps such as Canva and Spotify without leaving its interface. The San Francisco-based artificial intelligence (AI) firm also introduced AgentKit, a developer-focused tool to build, deploy, and optimise agentic workflows. Additionally, the company also made GPT-5 Pro, Sora 2, and Sora 2 Pro available via application programming interface (API).
ChatGPT Apps: Building on ChatGPT's agentic capabilities, OpenAI is now allowing software and services to connect to a new ChatGPT software development kit (SDK), which will then allow users to access these apps from the chatbot's interface. It will also let them complete certain tasks in these apps directly via text prompts.
Currently, users will be able to use Booking.com, Canva, Coursera, Expedia, Figma, Spotify, and Zillow in the chatbot. However, in the coming weeks, the company plans to add support for more apps such as DoorDash, OpenTable, Target, and Uber.
Apps SDK: Alongside ChatGPT apps, the company also released an open standard built on Anthropic's Model Context Protocol (MCP), dubbed Apps SDK. This will allow developers to build and test apps in ChatGPT. App submissions for publication will occur later this year.
OpenAI also plans to provide users with a directory to browse these ChatGPT-based apps. The company also promised to share details about the monetisation of these apps soon. Notably, the monetisation could also pave the way for ads on ChatGPT, which was reported last month.
Sora 2 and Sora 2 Pro in API: OpenAI has also made the Sora 2 and Sora 2 Pro AI models available via API. Sora 2 can generate videos in 1280x720p resolution, while Sora 2 Pro offers a resolution of up to 1792x1024p. Both models are capable of generating both landscape and portrait videos of up to 12 seconds duration. However, video input and image-to-video of real people is not supported at this time.
Notably, there are three variants available. The Sora 2 model is available at the price of $0.1 (roughly Rs. 8.8) per second of generated video, while the base Sora 2 Pro will cost developers $0.3 (roughly Rs. 26.6) for the same duration. Both of these models will generate videos with a watermark. To get watermark-free videos, devs will have to opt for the advanced Sora 2 Pro model, which charges $0.5 (roughly Rs. 44.3) per second of generation.
GPT-5 Pro in API: Alongside Sora, the GPT-5 Pro AI model is also available via API. The model uses more compute to think longer and provide more in-depth responses. The AI model is available in the Responses API and supports multi-turn interactions before responding to API requests. It is a fairly expensive offering with $15 (roughly Rs. 1,330) per million input tokens and $120 (roughly Rs. 10,650) for one million output tokens.
AgentKit for Developers: OpenAI describes AgentKit as a modular toolkit for building, deploying, and optimising agents. Developers can build an agent workflow, connect to large language models (LLMs), add tools support, add knowledge hubs, and even create voice-based agents.
Apart from this, Codex is now generally available, and the GPT-5 API requests are faster on the priority processing tier.
Catch the latest from the Consumer Electronics Show on Gadgets 360, at our CES 2026 hub.