OpenAI unveiled a new all-inclusive toolkit on Monday that is said to greatly facilitate the process of developing and deploying AI agents among developers. It was unveiled on October 6, 2025, when the company was hosting the Dev Day event, with the highlight being AgentKit among some other features aimed at improving the experience on ChatGPT.
The AI powerhouse is solving something that many developers have complained about as a significant pain point within the industry. Creating workable AI agents previously was a disjointed, time-consuming endeavor that saw developers dealing with complicated orchestration systems, having to handcraft custom connectors from the ground up, setting up the evaluation pipelines manually, and spending days on frontend development before something could be shipped.
“Historically, building agents involved wrangling disjointed tools, advanced orchestration with no version control, one-off connectors, eval pipelines by hand, prompt fine-tuning, and frontend work by the week before release,” OpenAI detailed in a corporate blog post. “With AgentKit, developers can design workflows in the browser and integrate agentic UIs quickly.”
OpenAI Launches Agent Builder (Beta), ChatKit, and Evals for Agents
The new toolkit packages multiple robust components into one package. Agent Builder enables developers to create AI agent workflows through a visual interface, stripping out much of the technical overhead that had tended to hold things back before. Because the process is visual, teams can prototype and refine agent behavior much more rapidly than before.
The Connector Registry provides centralized control over data and tools of the suite of products from OpenAI. It solves one common problem that comes with having multiple AI implementations in the same organization knowing what is connected to what, and that the appropriate access controls are in force.
Another notable new addition is the new ChatKit, which will allow developers to integrate custom agentic chat into their products. This, in turn, will allow businesses to create custom, branded AI interactions without having to design everything from scratch.
Perhaps most importantly for quality assurance, the toolkit includes Evals for Agents with enhanced capabilities. These evaluation tools come with features like datasets for testing, trace grading to understand agent decision-making, automated prompt optimization, and support for third-party models.
These features help developers measure how well their AI agents perform and identify areas for improvement.
OpenAI Tool Releases: Beta, GA, and Enterprise Access
OpenAI is rolling out these tools on a staggered schedule. Agent Builder is still in beta, allowing the early birds the opportunity to try the visual workflow designer and give them some feedback before making the tool generally available.
ChatKit and the new evaluation features have already gone generally available, allowing any developer who wants access to get on board immediately.
The Connector Registry is being deployed more selectively. OpenAI is currently providing beta access to select API users along with customers of ChatGPT Enterprise and Education.
This selective release is likely due to the enterprise-grade quality of centralized connector management and the desire to ensure that it works well with existing infrastructural hardware in the organization.
OpenAI’s Pivot to Actionable AI and Developer Viability
The Dev Day event wasn’t just about AgentKit. OpenAI also announced the ability to access apps directly within ChatGPT, expanding the platform’s functionality beyond conversation into actionable integrations. This move continues the company’s push to make ChatGPT a more versatile tool that can handle complex, multi-step tasks across different applications.
The release of the AgentKit suggests that OpenAI has finally agreed that the future of AI isn’t necessarily about better models, but rather making the models usable and viable for developers to leverage. By reducing the technical barrier and time required to work on AI agents, OpenAI is potentially leading the way towards a new set of AI-powered applications across sectors.
This can be quicker rollout of customer service bots, internal productivity, and custom AI assistants that are specific to certain workflows for businesses. For developers, this translates into fewer hours spent struggling with the infrastructure and more time spent on building meaningful AI experiences.
As AI agents become more intelligent and capable of performing complex tasks, software like AgentKit may be vital in determining the companies that will be able, at scale, to leverage such technology.
