内容来源：https://aibusiness.com/agentic-ai/openai-agentkit-enables-agentic-integration-in-chat

内容总结：

近日，OpenAI正式推出名为AgentKit的全新工具套件，旨在帮助开发者及企业用户通过聊天界面构建、部署和优化多智能体工作流。这一举措被行业观察者视为该公司在快速发展的智能体AI市场中确立领导地位的关键战略布局。

10月6日发布的AgentKit包含三大核心组件：用于创建和版本化管理多智能体工作流程的“智能体构建器”、统一管理工具与数据连接的“连接器注册中心”，以及可嵌入定制化聊天交互体验的“聊天工具包”。同时，平台还升级了模型评估体系，新增数据集管理、工作流自动诊断、提示词优化等功能，并开放强化学习微调服务的部分新能力。

行业分析师指出，此举显著降低了开发门槛。GAI Insights首席执行官保罗·贝尔强调：“这实际上将OpenAI定位为软件开发平台，类似苹果应用商店的生态模式。”他进一步表示，这种深度集成ChatGPT的策略，既提升了开发效率与安全性，也可能改变用户获取信息的习惯——越来越多用户正绕过传统搜索引擎直接使用ChatGPT。

在全球化竞争层面，Omdia分析师苏炼基指出，西方AI厂商中OpenAI率先实现了聊天界面与智能体能力的深度融合。虽然中国字节跳动旗下的“多宝”等产品已具备类似原生集成能力，但其生态相对封闭。而OpenAI的突破在于通过统一接口整合了从Codex编程工具到API服务的全生态资源。

面对行业竞争，谷歌与Anthropic等厂商也已布局智能体AI领域。谷歌于7月在Gemini模型中推出商务通话功能，Anthropic最新模型则为智能体AI奠定基础。但专家认为，OpenAI凭借ChatGPT的庞大用户基础和完整工具链，已建立起显著竞争优势，这或将对后续入局者构成更高壁垒。

（本文基于公开技术资料及行业分析师观点综合整理）

中文翻译：

由谷歌云赞助
选择您的首个生成式AI应用场景
想要入门生成式AI，首先应关注能够优化人机信息交互体验的领域。

该工具包旨在通过聊天界面创建并管理多智能体工作流。
OpenAI最新推出的AgentKit为开发者和商业用户提供了构建、部署、优化AI智能体并将其集成至聊天界面的新机遇。
这家生成式AI供应商于10月6日发布全新工具套件，包含面向开发者与商业用户的功能模块：Agent Builder、Connector Registry及ChatKit。

工具包与新功能解析
OpenAI将Agent Builder定义为创建多智能体工作流并进行版本管理的操作平台。
Connector Registry是管理员统一管理OpenAI产品间工具与数据连接的核心枢纽。ChatKit则支持用户将可定制的聊天式智能体体验嵌入自有产品。
OpenAI同时宣布为评估平台evals新增四项功能，助力用户测评大语言模型表现。
新增评估功能包含支持用户构建智能体评估的数据集，并可通过人工标注与自动评分实现持续扩展。
Trace grading功能可自动识别并修复工作流缺陷。自动提示优化则能基于人工标注与评分结果生成提示词。
在强化微调（RFT）方面，OpenAI推出新功能支持开发者定制推理模型。RFT已在OpenAI o4-mini全面开放，GPT-5版本开启小范围测试。测试版新增自定义工具调用功能，据OpenAI介绍，该功能可训练模型在恰当时机调用工具，从而提升推理能力。

战略布局彰显行业领导力
观察人士指出，AgentKit是OpenAI的重要战略举措，彰显其在快速发展的智能体AI市场中构建差异化的雄心。
“这是关键性的一步，”GAI Insights首席执行官保罗·贝尔表示，“此举旨在将OpenAI打造成供其他软件公司开发应用的平台。”
他补充说，这或将使OpenAI成为“智能体应用商店”，类似苹果的App Store模式。
其重要意义在于降低开发门槛，推动更多应用以ChatGPT而非浏览器作为交互界面。
“这将促使更多企业开发从ChatGPT到自身产品的扩展功能，反之亦然，”贝尔分析道，“这种模式更便捷、成本更低、安全性更高，有利于OpenAI构建生态壁垒，并对Chrome等浏览器构成挑战。”
他同时指出，ChatChatGPT已积聚极高人气，大量用户绕过谷歌搜索直接使用该服务。

Informa TechTarget旗下机构Omdia分析师苏建熹表示，将智能体能力整合进聊天界面，也显示出OpenAI相对于部分中国厂商的领先优势。
“这可能是西方世界首个将聊天界面与智能体AI能力相融合的生成式AI供应商案例，”苏建熹指出。
他提到中国科技巨头字节跳动已在旗下聊天机器人“多宝”中集成类似AI通话功能，但强调该功能目前仅限字节跳动生态内使用，未开放第三方软件支持。
“字节跳动在原生智能体AI集成与开发平台方面领先OpenAI，但OpenAI通过聊天界面实现无缝集成的能力冠绝业界，”苏建熹评价道。

优势、挑战与竞争对手
OpenAI的美国竞争对手们也正将生成式AI平台向智能体AI方向演进。谷歌已于7月通过Gemini模型推出商务通话功能，Anthropic最新模型Sonnet 4.5同样为智能体AI奠定基础。
苏建熹补充说，OpenAI通过AgentKit实现的集成允许用户不仅与模型交互，更能触达整个生态体系，包括Codex、API乃至ChatGPT等开发者工具。
“最显著的优势在于，用户无需在不同平台间辗转，通过单一界面即可完成所有操作，”他解释道，“鉴于ChatChatGPT本就是最受欢迎的交互界面之一，AgentKit用户如今既能享用丰富工具，又拥有多样化选择。”
苏建熹认为AgentKit给其他生成式AI供应商带来的挑战在于，智能体AI市场的准入门槛将被进一步抬高。

OpenAI还推出了新一代应用程序，用户可在ChatGPT内直接与之对话。该公司表示开发者可通过新发布的预览版Apps SDK开始构建应用。

您可能还喜欢

英文来源：

Sponsored by Google Cloud
Choosing Your First Generative AI Use Cases
To get started with generative AI, first focus on areas that can improve human experiences with information.
The toolkit is aimed at creating and managing multi-agent workflows from a chat interface.
OpenAI's new AgentKit provides developers and business users with a new opportunity to build, deploy, and optimize AI agents and integrate them into a chat interface.
The generative AI vendor launched the new tool suite on Oct. 6. The suite comprises developer and business user-oriented capabilities, including Agent Builder, Connector Registry and ChatKit.
The kit and other new features
OpenAI refers to Agent Builder as a canvas for creating and versioning multi-agent workflows.
Connector Registry is a hub where admins can manage how tools and data connect across OpenAI products. ChatKit enables users to embed customizable chat-based agent experiences in their products.
OpenAI also revealed that it has added four new capabilities to evals, which enable users to evaluate and test a large language model's performance.
The new eval capabilities include datasets that let users build agent evals and expand them over time with human annotations and automated graders.
Another new capability is Trace grading, which enables users to identify and address workflow shortcomings automatically. Automated prompt optimization lets users generate prompts based on human annotations and grader outputs.
Along with new evaluations, OpenAI introduced new features to reinforcement fine-tuning (RFT), which enables developers to customize reasoning models. RFT is generally available on OpenAI o4-mini and in private beta for GPT-5. New capabilities in beta include custom tool calls that enable users to train models so they use the right at the right time, resulting in improved reasoning, according to OpenAI.
Show of leadership
AgentKit is an important strategic move for OpenAI that shows how it is trying to differentiate itself in the fast-moving agentic AI market, observers said.
“It’s a really important play,” said Paul Baier, CEO of GAI Insights. “It’s really positioning OpenAI to be a platform for other software companies to develop software within there.”
He added that this move can position OpenAI as an agent store, similar to the way Apple has an Apple store.
It is important because it makes it easier for developers to build applications that use OpenAI's ChatGPT as an interface instead of a browser.
“That’s going to allow more companies to develop extensions from OpenAI ChatGPT to their products, and vice versa,” Baier said. “It’s easier, less costly, more secure, and that bodes really well for lock-in for OpenAI, and it will make it more challenging for things like Chrome.”
He added that OpenAI’s ChatGPT is immensely popular and many users bypass Google Search to go to ChatGPT.
Integrating agentic capability within a chat interface also shows how OpenAI is ahead of some Chinese vendors, according to Lian Jye Su, an analyst at Omdia, a division of Informa TechTarget.
"This is probably the first instance in the Western world ... where a GenAI vendor is starting to integrate their chat interface with agentic AI capability," Su said.
He noted that Chinese tech giant ByteDance already has similar features that integrate AI chat calling within its chatbot, Duobao. However, Duobao's agentic AI call ability is limited to the ByteDance ecosystem and does not support third-party software, he continued.
"ByteDance is ahead of OpenAI in native agentic AI integration and development platform offerings, but OpenAI is ahead of everyone in the seamless integration through chat interface," Su said.
Benefits, challenges and rivals
But some of OpenAI's U.S. competitors are also moving their generative AI model platforms toward agentic AI. Google has rolled out a similar feature with its Gemini model's business calling feature in July. Anthropic's latest model, Sonnet 4.5, also sets the groundwork for agentic AI.
Su added that OpenAI's integration with its AgentKit lets users interact not just with OpenAI's models but across the entire OpenAI ecosystem, which includes developer tools such as Codex, its API and even ChatGPT.
"The clear benefit is that instead of having to go through all these different hoops to integrate everything, they can now do everything under a single interface," Su said. He added that since the ChatGPT interface is already one of the most popular interfaces available, users of AgentKit now have access to both a wide range of tools and a broad range of options.
The challenge posed by AgentKit is for other generative AI vendors is they might now find it harder to enter the agentic AI market, Su said.
OpenAI also introduced a new generation of apps that users can chat with inside ChatGPT. The vendor said developers can start building apps with a new Apps SDK in preview.
You May Also Like