Openai reveals the API of the answers, Open Code SDK agents allow developers to build their own deep research and operator

Openai is Deploying a new set of APIs and instruments Designed to help developers and businesses build more efficiently agents driven by AI Deep research (which interferes with the Internet independently of developing richly studied, well -organized and cited reports) and Operator (His web browser control tool autonomously based on the text instructions of the user and performing actions such as finding sports tickets or reservations).

Now, with access to the building blocks behind these powerful OPENAI agents from the first country, developers can build their own competitors in third parties or more special products and services specific to their case and audience.

Openai’s last progress in reflection mechanisms, multimodal processing and safety laid the foundations for these capabilities, especially its “O” family of reasoning models (O1 and O3).

“It is difficult to overestimate how critical the models of AI agents activate are,” says Olivier God -God, the leader of the Openai Platform Product, in VentureBeat video call. “One of the biggest restrictions before was to deal with long horizon tasks like planning.”

But the company says that developers have not yet had the tools needed for easily built them in applications ready for manufacturing for businesses and their customers to this day.

To deal with these obstacles, Openai introduces several new suggestions: API of answers, built -in network search tools and files, a computer use tool and open source SDK agents. While the API of the answers allows developers to build agents at the top of their technologies, SDK agents can help them connect agents to other web instruments and processes by performing “work flows” that do what the user or business wants autonomously.

These tools aim to optimize the development of AI agents by reducing the need for extensive logic of engineering and personalized orchestration. They should also Manus., Qwen of Alibaba., Deepseekand domestic rivals like Anthrop and GoogleS

While these other players offer tools or products to developers, the continuous development of the OpenAi developer platform makes a difficult offer to beat as a “stop shop” for those who want to use the latest AI advances in a clean, easy and affordable way.

In progress to send AI blogosphere and space on social media, Openai returns to an open code in a big way with the launch of its SDK agents, an instruments designed to help developers manage, coordinate and optimize the work processes of agent-based agents, which are powered by other models. Anthropica and Google, or open source models from other models, such as competitors, anthropic and Google, or models of OpenA family.

“The SDK agent has an open source that allows businesses to mix and compare different models,” said Godment. “We don’t want to force anyone to use only Openai models.”

SDK offers key features such as:

• Configurable agents – AI models with predefined instructions and access to tools.

• Intelligent gears – mechanisms for transferring tasks between agents based on context.

• Built -in railings – safety measures to validate input and moderation of content.

• Tracking and observation – Tools for debugging and optimizing the productivity of the agent.

“With SDK agents, developers can track what exactly the agent is doing – what tasks, that it arises, what data it collects and how it generates answers,” says Nikun Handa, PM in Openai’s API team, in the same Venturebeat video call.

What does API offer for new answers

At the center of this update is the API of the answers, which combines the API characteristics to complete the Openai chat with the functionality of using the Assistants API tool, the last of which will be withdrawn in the mid -2026, According to the companyS

This integration allows developers to use multiple built -in tools within a single API call, which facilitates the construction of applications that require complex, multi -stage interactions.

API for answers initially supports three built -in instruments:

• Web Search -The states cited answers in real time by retrieving information from the network.

• Searching a file – extracts the relevant information from large document storage facilities using filtration of metadata and optimized request processing.

• A tool to use the computer – Allows AI agents to take action on a computer, such as surfing, entering data and navigating software interfaces.

“With API to the answers, the developers get more visibility in what the model does – what tools are calling, why he calls them and what decisions he makes before and after these calls,” Handa said.

With these capabilities, Openai provides API of the answers that serve as the basis for agent applications, eliminating the need for multiple external integrations. API is available to all developers today, using the use of standard Openai markers and tools.

In addition, Openai notes that while the API to complete the chat will continue to receive updates, the API of the answers is considered its own superset. Developers who need built -in tools or multi -stage interpretations of the model should use API for answers for new integration.

Openai also makes its network search tools, file search and computer use on a computer, available directly through API for answers. These tools allow AI agents to have access to real information, to extract context from documents and to interact with digital environments more efficiently.

Web search offers real -time developers with quotes

The network search tool allows developers to integrate real -time search opportunities into their applications, making it useful for assistants, shopping guides, and content aggregation tools. It provides sources for its answers, ensuring that users can check the accuracy of the information.

“The first thing we start are built -in tools, such as network search, allowing models to access real -time information,” Handa said. “This is the same tool that possesses Chatgpt’s demand, and now we bring it to API.”

Openai also confirmed that the search results on the API network will include clear quotes, allowing users to click on original sources. Developers can introduce web search as part of a width extraction system that includes its own data sources.

File search: Intelligent retrieval of private cloud documents

With file search tool, AI agents can quickly retrieve the relevant information from large collections of documents. This tool supports multiple file formats and includes features such as request optimization, metadata filtration and custom ranking for more accurate results.

“The third tool we launch is the search for files that make it easier for developers to take all their data, store it in our system and retrieve the right information with high accuracy,” Handa explained.

The file search tool is priced at $ 2.50 per thousand of requests, with storage fees being $ 0.10 a day (the first GB is free).

Developers can now gain access to computer use, Openai’s Teching Openai operator

The computer use tool extends the agent’s capabilities beyond simple text tasks, allowing AI to interact with computer interfaces.

Powered by the OPENAI AGENT-AGENT (CUA) computer use model, this tool translates AI-generated actions into executable commands, which allows automation of tasks such as data entry and web navigation.

“We are also launching a computer use tool that allows models to interact with graphic user interfaces when there is no existing API,” Handa noted.

Currently, the computer use tool is available as a review of research for selected developers at 3-5 use levels. The pricing is defined at $ 3 per million input tokens and $ 12 per million markers.

What it means to businesses of enterprises

For IT team leaders, CTOS and mid-level managers who want to optimize work processes, the new Openai tools provide a clear path to automation and scaling of AI-moving processes without requiring a widespread development.

The built -in file search and search capabilities allow businesses to quickly integrate the retrieval of information powered by AI into its existing systems, while the computer use tool allows automated interactions with inherited applications that do not access API.

Open source SDK agents further authorize organizations to coordinate working processes managed by AI to facilitate the deployment of agents that improve efficiency in areas such as customer support, document processing and market research.

With the security and observability of the enterprise built into these instruments, the decision-making persons can adopt AI decisions with greater transparency and control, ensuring that compliance and monitoring of efficiency on a scale.

What is next?

Openai sees these new editions as the first step in the construction of a comprehensive AI agent platform. The company plans to unleash additional tools and integrations in the coming months to help developers deploy, evaluate and score more effectively agency applications.

“We believe that the coming months will be crucial to the deployment of more and more agents on a scale,” Bozic said. “We have already done this with agents on the first countries like Deep Research, but Openai will not build every agent-so we have a developer platform.”

Openai also said it would continue to improve the safety functions of agency applications, including protective measures for quick injections and unauthorized data access.

Developers who are interested in building with new instruments can investigate the API’s Openai and Playground documentation to start today.