OpenAI's new Agent API toolbox
Created on March 12|Last edited on March 12
Comment
OpenAI has introduced a new set of APIs and tools designed to help developers create AI agents that can complete tasks independently. These updates aim to streamline development by reducing the need for complex prompt engineering and manual orchestration.
Responses API: A More Powerful Way to Build AI Agents
The new Responses API combines features from the Chat Completions and Assistants APIs, making it easier for developers to integrate AI models into their applications. It supports built-in tools like web search, file search, and computer use, allowing AI agents to interact with real-world data and complete multi-step tasks with a single API call.
Web Search for Real-Time Information
A key addition is the built-in web search tool, which allows AI agents to fetch up-to-date information from the internet with proper citations. This feature is particularly useful for research assistants, shopping bots, and travel planners. The same search model is also available separately in the Chat Completions API, providing developers with access to OpenAI’s fine-tuned search capabilities.
File Search for Retrieving Documents
The new file search tool helps AI agents quickly find relevant information within large document collections. It supports multiple file types, offers metadata filtering, and includes query optimization to improve accuracy. Businesses can use this tool for customer support, legal research, or any task that requires retrieving specific information from stored documents.
Computer Use for Automating Tasks
OpenAI has introduced a computer use tool that enables AI agents to interact with digital interfaces, such as filling out online forms, navigating websites, or performing repetitive tasks on a computer. This tool translates user requests into keyboard and mouse actions, making it useful for automating workflows in industries that rely on legacy systems. While still in research preview, it has already been used in enterprise applications to streamline operations.
Agents SDK for Multi-Agent Workflow Management
In addition to new tools, OpenAI has released an open-source Agents SDK to help developers orchestrate multi-agent workflows. This SDK makes it easier to build AI agents with predefined roles, manage interactions between them, and implement safety measures. Companies like Coinbase and Box have already used the SDK to deploy AI-powered assistants that help with research, customer support, and data processing.
Costs and Availability
These new tools are available to developers, but they are not free. The Responses API follows OpenAI’s standard token pricing, and individual tools have their own costs. Web search starts at $30 per 1,000 queries, file search is $2.50 per 1,000 queries, and the computer use tool costs $3 per 1 million input tokens and $12 per 1 million output tokens. The Agents SDK is free, but using OpenAI’s models and tools will still incur usage fees.
What’s Next for AI Agents?
OpenAI plans to continue expanding its platform with more integrations and advanced features for AI agents. As AI becomes more capable, OpenAI aims to provide businesses and developers with the tools needed to build autonomous systems that can handle increasingly complex tasks. These updates mark a major step toward making AI-powered agents a practical and scalable solution for real-world applications.
Add a comment
Tags: ML News
Iterate on AI agents and models faster. Try Weights & Biases today.