OpenAI Launches New Tool to Help Developers Build AI Agents

While rolling out ChatGPT paid subscription plans to individual users brings in significant revenue, OpenAI's commitment to supporting the AI ​​developer community remains strong.

Viewing AI agents as the future of the artificial intelligence industry, OpenAI acknowledges that building robust, “industrial-grade” agents with current tools is difficult. To address this, the company has just announced a new toolkit that simplifies the creation of advanced AI agents capable of handling complex, multi-stage operations for both developers and enterprises.

Responses API: New Tools for Developers

Responses API is a new “primitive” API that allows developers to leverage OpenAI’s built-in tools to build their own agents. Responses API combines the simplicity of Chat Completions with the tooling power of Assistants API. With a single Responses API call, developers can perform complex tasks using multiple tools, including web search, file search, and calculator, and handle multiple interactions with the model.

With the Responses API, developers can also easily store data to evaluate agent performance with features like tracing and evaluations. The Responses API will replace the Assistants API in mid-2026.

Responses API comes with the following built-in tools:

1. Web Search: Allows developers to get up-to-date answers with relevant citations from the web. The new web search engine can be used with gpt-4o and gpt-4o-mini. OpenAI notes that the GPT‑4o search preview and GPT‑4o mini search preview scored 90% and 88%, respectively, on the SimpleQA test. The web search engine costs $30 and $25 per thousand queries for GPT‑4o search and 4o-mini search.
2. File Search: Allows developers to easily find relevant information from large volumes of files. The new tool supports multiple file types, query optimization, metadata filtering, and custom reordering. File Search costs $2.50 per 1K queries, and file storage is $0.10/GB/day, with the first 1GB free.
3. Computer-Using: Allows developers to complete tasks on computers. This tool is powered by the same Computer-Using Agent (CUA) model that powers the OpenAI Operator. Computer-Using costs $3/1 million tokens input and $12/1 million tokens output.

Agents SDK: Multi-Agent workflow support

In addition to the new underlying tools, OpenAI is also announcing a new Agents SDK to help developers coordinate multi-agent workflows. It comes with improvements over the Swarm experimental agents SDK that OpenAI released last year:

• Agents: LLMs are easy to configure with clear instructions and built-in tools.
• Handoffs: Intelligent transfer of control between agents.
• Guardrails: Configurable security checks to validate input and output.
• Tracing & Observability: Visualize agent execution for debugging and performance optimization.

The Agents SDK will work with the Responses API and Chat Completions API. Developers can also use the same SDK with non-OpenAI AI models if they support the Chat Completions API endpoint.

OpenAI is committed to enhancing its platform with deeper API integrations and new tools for deploying, evaluating, and optimizing agents, especially as models with improved agentic capabilities emerge.