OpenAI’s newly launched AgentKit offers developers and business users fresh opportunities to build, deploy, and optimize AI agents, seamlessly integrating them into chat interfaces. The generative AI provider unveiled the new toolkit on October 6, 2025, introducing features tailored for both developers and business professionals. The toolkit includes Agent Builder, Connector Registry, and ChatKit. **New Tools and Features** Agent Builder functions as a canvas for creating and managing multi-agent workflows, according to OpenAI. The Connector Registry acts as a centralized hub where administrators can manage how tools and data connect across OpenAI products. ChatKit enables users to embed customizable, chat-based agent experiences directly into their applications. OpenAI also enhanced its Evals platform with four new features designed to help users assess and test the performance of large language models. These new evaluation tools include datasets for building agent evaluations and scaling over time using human annotations and automated scoring systems. Another feature, trace grading, allows users to automatically detect and resolve workflow flaws. Auto-prompt optimization enables users to generate prompts based on human feedback and scorer outputs. **Introducing Reinforcement Micro-Tuning (RFT)** In addition to the new evaluation tools, OpenAI introduced Reinforcement Fine-Tuning (RFT), allowing developers to customize reasoning models. RFT is generally available on the OpenAI o4-mini model and currently in beta testing for GPT-5. According to OpenAI, beta features include custom tool calling, enabling users to train models to make optimal tool usage decisions, enhancing reasoning performance. **Demonstrating Strategic Leadership** Industry analysts believe AgentKit represents a major strategic move by OpenAI to differentiate itself in the fast-evolving AI agent market. “This is a very significant move,” said Paul Baier, CEO of GAI Insights. “It positions OpenAI as a platform on which other software companies can build internally.” He added that the move could position OpenAI as an agent marketplace, similar to Apple’s App Store. This development makes it easier for developers to create apps using OpenAI’s ChatGPT as an interface instead of a browser. “It will enable more companies to develop extensions from OpenAI ChatGPT into their products—and vice versa,” Baier said. “It’s easier, cheaper, and more secure, which is a positive sign for OpenAI’s ecosystem lock-in, potentially challenging products like Chrome.” He noted that ChatGPT is already so popular that many users are bypassing Google search in favor of ChatGPT. Lian Jye Su, an analyst at Omdia, part of Informa TechTarget, said that integrating agent capabilities into chat interfaces also highlights OpenAI’s leadership over some Chinese AI vendors. “This may be the first time in the Western world that a GenAI vendor has started integrating chat interfaces with agent AI features,” Su stated. He pointed out that Chinese tech giant ByteDance already offers similar functionality, embedding AI chat calling into its chatbot, Doubao. However, Doubao’s agent AI calling is limited to the ByteDance ecosystem and does not support third-party software. “ByteDance leads OpenAI in native agent AI integration and development platform products, but OpenAI is ahead of everyone in seamless integration via chat interfaces,” Su added. **Competitive Landscape and Challenges** However, some of OpenAI’s U.S. rivals are also transitioning their generative AI platforms toward agent AI. Google launched similar functionality in July with its Gemini model’s business calling feature. Anthropic’s latest model, Sonnet 4.5, also lays the foundation for agent AI. Su noted that OpenAI’s AgentKit integration allows users not only to interact with OpenAI models but also with the entire OpenAI ecosystem, including tools like Codex, its API, and even ChatGPT. “The clear advantage is that users can now do everything under one interface, without having to go through multiple integrations,” Su said. He added that since the ChatGPT interface is already one of the most popular, AgentKit users now have access to a wide range of tools and options. He also pointed out that AgentKit presents a challenge for other generative AI providers, as entering the agent AI market may now become more difficult. OpenAI has also launched a new generation of apps that users can interact with directly inside ChatGPT. The company stated that developers can now start building apps using the preview version of the new app SDK.