OpenAI has launched an agent creation tool that the company hopes will eliminate fragmented tooling and make it easier for enterprises to apply OpenAI to create agents.
Agent kitannounced at OpenAI DevDay in San Francisco, allows developers and enterprises to create agents and add chat capabilities in one place, potentially competing with platforms like Zapier.
By offering a more streamlined way to create agents, OpenAI is expanding further to become a full-stack application provider.
“Until now, software development agents meant juggling fragmented tools — complex orchestration without versioning, custom connectors, manual evaluation pipelines, rapid tuning, and weeks of pre-launch UI work,” the company said in a blog post.
AgentKit includes:
-
Agent Builder, a visual canvas where developers can see what they’ve built and version multi-agent workflows
-
The Connector Registry is a central area where administrators can manage connections between OpenAI products. The Global Admin Console will be a prerequisite for using this feature.
-
ChatKit enables users to integrate chat-based agents with user interfaces.
Ultimately, OpenAI said it would build a standalone Workflows API and add agent deployment tabs to ChatGPT.
OpenAI has also expanded agent scoring, adding capabilities such as datasets with automatic scoring and annotation tools, trace scoring that performs end-to-end assessment of workflows, automatic tooltip optimization, and support for third-party agent measurement tools.
Developers can access some of AgentKit’s features, but OpenAI is gradually rolling out additional features like Agent Builder. Currently, Agent Builder is available in beta, with ChatKit and modern evaluation features generally available. Connector Registry is “beginning beta rollout to some API and ChatGPT Enterprise and Edu users.
OpenAI says AgentKit pricing will be included in pricing for standard API models.
Agent Builder
To be clear, many agents are built using OpenAI models; however, enterprises often access GPT-5 through other platforms to create their own agents. However, AgentKit brings enterprises more into its ecosystem so they don’t have to use other platforms as often.
Demonstrated at DevDay, the company found Agent Builder ideal for rapid iteration. It also provides developers with insight into how agents operate.
In the demo, an OpenAI developer created an agent that reads the DevDay agenda and suggests panels to watch. It took her less than eight minutes.
Other model providers have recognized the importance of offering agent development toolkits to encourage enterprises to use their tools more often. Google in April, it released the Agent Development Kit, expanding the ability to create multi-agent systems “in less than 100 lines of code.” Microsoftwhich powers the popular AutoGen agent platform, announced that its new Agent Framework will make agent creation one place.
OpenAI customers such as fintech company Ramp said in a blog post that its teams were able to build a purchasing agent in hours instead of months.
“Agent Builder has transformed what once required months of convoluted orchestration, custom code and manual optimizations in just a few hours. The visual canvas keeps product, legal and engineering at the same level, shortening iteration cycles by 70% and launching the agent in two sprints rather than two quarters,” Ramp said.
The AgentKit connector registry would also enable enterprises to manage and maintain data across different workspaces, consolidating data sources into a single dashboard covering both ChatGPT and API. It will have ready-made connectors for Dropbox, Google Drive, SharePoint and Microsoft Teams. It also supports third-party MCP servers.
Another capability of Agent Builder is Guardrails, an open-source security layer that protects against personal data leakage, jailbreaking, and unintended or malicious behavior.
More chat
Since most agent interactions involve chat, it makes sense to simplify the process for developers to configure chat interfaces and connect them to the agents they just created.
“Implementing chat interfaces for agents can be surprisingly complex – handling streaming responses, managing threads, demonstrating model thinking, and designing engaging chat experiences,” OpenAI said.
The company says ChatKit makes it uncomplicated to embed chat agents on platforms and embed them in apps or websites.
However, some of OpenAI’s competitors have begun to think beyond the chatbot and want to offer agent-based interactions that are more seamless. Google’s asynchronous encoding agent, Jules, has introduced a modern feature that allows users to interact with the agent via a command-line interface, eliminating the need to open a chat window.
Answers
Reaction to AgentKit has been mostly positive, with some developers noting that while it simplifies agent creation, it does not mean that anyone can now create agents.
Several developers see Agent Kit not as a Zapier killer, but rather as a pipeline complementary tool.
Zapier debuted in 2024 with a no-code tool for creating AI agents and bots called Zapier Central.