On the first anniversary of GPT-5, OpenAI launches the Agent Plugins standard: ending the fragmentation of agent plug-ins and defining cross-client interoperability standards. On the first anniversary of the launch of the GPT-5 series models, OpenAI announced the launch of Agent Plugins, a set of plug-ins for AI agents.

On the first anniversary of GPT-5, OpenAI launches the Agent Plugins standard: ending the fragmentation of agent plug-ins and defining cross-client interoperability standards. On the first anniversary of the launch of the GPT-5 series models, OpenAI announced the launch of Agent Plugins, a set of plug-in packaging standards for AI agents. The standard is positioned as an open, vendor-neutral specification that aims to unify the way to extend the capabilities of AI agents by packaging reusable components into portable plug-ins. Currently, version 1.0.0 of the specification has been officially launched, defining a shared format covering Agent Skills and MCP Servers. Compatible clients can discover and load these components according to the same rules, without the need to repeatedly adapt for different platforms. In the announcement, OpenAI pointed out the core pain point of the current AI agent ecosystem: different clients have formed their own independent plug-in formats. Even if the plug-ins contain the same underlying components, developers often need to rearrange or copy components for different clients. Plug-ins packaged on one client often require additional adaptation before they can be used on another client, which results in low plug-in development efficiency and serious duplication of work. Agent Plugins attempt to set a lightweight interoperability foundation for parts that are portable across clients, with shared components using a predictable directory structure, while client-specific capabilities such as distribution, installation, permissions, and user experience continue to be controlled by each client. Unified directory structure reduces development threshold According to specifications, an Agent Plugin directory must contain the manifest file plugin.json, which is used to identify the plug-in and its target version. The skills directory contains skill components defined by the Agent Skills specification, and mcp.json describes stdio, Streamable HTTP or traditional HTTP+SSE type MCP servers. In addition, it also supports client-specific hooks directories to implement customized extensions. This structured directory design allows plug-in authors to write components only once and they can be discovered and loaded in all compatible clients, significantly reducing cross-platform development and maintenance costs. The launch of Agent Plugins marks that the AI ​​agent ecosystem is moving from individual operations to standardized collaboration. As the MCP protocol is adopted by more and more manufacturers, tool calling and capability sharing between agents has become an industry consensus, and the unification of plug-in formats is a natural extension of this trend. OpenAI chose to release this specification on the anniversary node of GPT-5, which is a response to the rapid development of the intelligent ecosystem in the past year.