AI
OpenAI launches Agent Builder tool
AgentKit is OpenAI’s latest play to win over developers in the race against Anthropic, Google, and others trying to make AI agents that handle the boring stuff.

Just a heads up, if you buy something through our links, we may get a small share of the sale. It’s one of the ways we keep the lights on here. Click here for more.
At OpenAI’s Dev Day 2025, CEO Sam Altman took the stage with his trademark grin and a brand-new toy for developers: AgentKit, a toolkit built to make creating AI agents as easy as building a website on Squarespace, but for robots that actually do stuff.
“AgentKit is everything you need to build, deploy, and optimize agent workflows with way less friction,” Altman said, pitching it like the Swiss Army knife of AI development.
The idea? Help developers go from half-baked prototypes to full-blown, autonomous agents that can take action, not just spit out text.
If last year’s theme was “ChatGPT can do more,” this year’s is “Now it can hire itself an assistant.”
AgentKit is OpenAI’s latest play to win over developers in a growing arms race against Anthropic, Google, and others trying to make AI agents that handle the boring stuff, think scheduling, data pulling, and decision-making, so you don’t have to.
The kit comes with a few shiny new parts. The star of the show is Agent Builder, which Altman likened to “Canva for agents.” (Via: TechCrunch)
It’s a drag-and-drop-style tool that lets developers visually design logic and steps without crying over API documentation.
Then there’s ChatKit, which gives anyone the power to embed chat interfaces directly into their own apps, branded, customized, and ready to charm users in your company’s preferred tone of voice.
For those who actually want to know if their agent is any good, Evals for Agents provides grading tools, datasets, and automated prompt optimization.
Basically, a report card for your robot coworker.
And Connector Registry lets developers hook up agents to internal tools and external systems safely, complete with an “admin control panel” that sounds suspiciously like a mission control for AI.
To show off, OpenAI engineer Christina Huang built two fully functional agents live on stage in under eight minutes.
The crowd loved it, and Altman quipped, “This is all the stuff we wished we had when building our first agents.”AgentKit isn’t just another dev tool. It’s OpenAI’s big bet that the next wave of apps won’t just talk, they’ll act.
Follow us on Flipboard, Google News, or Apple News
