DevDay 2025: OpenAI Launches AgentKit, Codex GPT‑5 & More; Know How To Use AI Agent Builder & In-Chat Apps

OpenAI has unveiled AgentKit, a comprehensive new toolkit designed to streamline the creation, deployment and optimisation of AI agents. The announcement was made by CEO Sam Altman during the DevDay conference, where he emphasised that the new suite seeks to simplify what has traditionally been a fragmented and complex development process.

DevDay 2025: Sam Altman Launches OpenAI's Agentkit, Codex & More

What is Agentkit?

AgentKit comprises several key components: Agent Builder, a visual drag‑and‑drop canvas to design multi‑agent workflows; ChatKit, which enables embedding customisable chat interfaces into applications and Evals, a performance evaluation module that supports trace grading, prompt optimisation and assessment of individual agent components.

The toolkit is built on top of OpenAI's existing Responses API, which many developers already use. AgentKit also integrates with a Connectors Registry, enabling agents to securely connect to internal business systems or third‑party services via an admin dashboard.

OpenAI says that ChatKit and the new Evals features are now generally available, while Agent Builder is in beta, and the Connector Registry is in limited beta for certain enterprise customers.

OpenAI Rolls Out Codex GPT‑5 and In‑Chat Apps Like Booking.com, Canva, Spotify, Zillow & More

Alongside the launch of AgentKit, OpenAI unveiled a suite of powerful new tools. One of the key highlights is the enhancement of Codex, now built on the GPT‑5 family, which can dynamically adjust its "thinking time" based on the complexity of tasks. This update is specifically optimised for agent-based coding workflows, enabling developers to build more intelligent and responsive systems.

OpenAI also announced the rollout of in-chat apps like Booking.com, Canva, Coursera, Expedia, Spotify, Figma, and Zillow, which now run directly within the ChatGPT interface.

How to Use OpenAI AgentKit: Check Step-by-Step Guide Here

Here is a step-by-step guide on how to use OpenAI's AgentKit to create and deploy your own AI agents as per OpenAI report.

• Log in to your OpenAI account

• Access AgentKit Dashboard: Visit the AgentKit dashboard or documentation page via the OpenAI platform. Ensure API access is enabled for AgentKit features. AgentKit access may require a Pro or Enterprise plan during the beta phase.

• Launch Agent Builder (Beta): Open the visual Agent Builder tool. Use its drag-and-drop interface to define agent goals, logic, and workflows.

• Configure Agent Logic: Set up specific tasks for your agent-like answering queries, making bookings, or fetching data. Define triggers, responses, and actions.

• Test and Refine with Trace Grading: Use the Trace Grading tool to analyse your agent's decision-making process step by step, improving reliability and performance.

• Deploy and Connect: Once your agent is ready, deploy it through the dashboard and integrate it with apps or tools using OpenAI's Connectors Registry.