Example use cases for Ontology MCP (OMCP)（本体论MCP（OMCP）的示例用例）¶
The following sections describe common usage patterns for Ontology MCP (OMCP).
Desktop agent integration¶
You can use Ontology MCP with desktop agents such as Claude.ai, Microsoft Copilot Studio, and Gemini Enterprise. In these use cases, Ontology MCP enriches the desktop agent with the ability to connect to your organization's source of truth, read and write data, and execute actions governed by your ontology.
This pattern enables human-agent symbiosis: both the human user and the AI agent operate on the same shared data, logic, actions, and governance. For example, a user working in Microsoft Teams can ask a Copilot Studio agent to create a task in the ontology, and the resulting data is immediately visible to other users and applications that share the same ontology.
External agentic workflow integration¶
Organizations that run agentic workflows using external frameworks such as Google Agent Development Kit (ADK), Microsoft Agent Framework, or OpenAI SDK can use Ontology MCP to access Foundry data and logic through a common interface. Agents in these frameworks interact with Ontology MCP the same way they interact with other systems, using MCP as a standard protocol.
Developers can combine multiple MCP servers to build richer agent capabilities. For example, an agent can use Perplexity MCP for web search alongside Ontology MCP for reading and writing ontology data, enabling workflows that span both external knowledge and your organization's source of truth.
Headless agents¶
Ontology MCP can serve as the memory and tracking layer for headless agent workflows. Headless agents operate without direct human supervision and can use Ontology MCP to both read and write from the organization's source of truth.
Because every action the agent performs through Ontology MCP is recorded in the ontology, this pattern provides auditability for unsupervised operations. You can use any agent framework, such as Anthropic SDK or Google ADK, to build headless workflows that run in response to ontology changes or on a scheduled basis.
中文翻译¶
本体论MCP（OMCP）的示例用例¶
以下章节描述了本体论MCP（Ontology MCP, OMCP）的常见使用模式。
桌面代理集成¶
您可以将本体论MCP与Claude.ai、Microsoft Copilot Studio和Gemini Enterprise等桌面代理配合使用。在这些用例中，本体论MCP增强了桌面代理的能力，使其能够连接到组织的真实数据源，读取和写入数据，并执行由本体论治理的操作。
这种模式实现了人机共生：人类用户和AI代理都基于相同的共享数据、逻辑、操作和治理规则进行工作。例如，在Microsoft Teams中工作的用户可以请求Copilot Studio代理在本体论中创建一个任务，生成的数据将立即对共享同一本体论的其他用户和应用程序可见。
外部代理工作流集成¶
使用外部框架（如Google Agent Development Kit (ADK)、Microsoft Agent Framework或OpenAI SDK）运行代理工作流的组织，可以通过本体论MCP使用通用接口访问Foundry数据和逻辑。这些框架中的代理与本体论MCP的交互方式与它们与其他系统的交互方式相同，均使用MCP作为标准协议。
开发人员可以组合多个MCP服务器来构建更丰富的代理能力。例如，一个代理可以使用Perplexity MCP进行网络搜索，同时使用本体论MCP读取和写入本体论数据，从而实现跨越外部知识和组织真实数据源的工作流。
无头代理¶
本体论MCP可以作为无头代理工作流的内存和跟踪层。无头代理在没有直接人工监督的情况下运行，可以使用本体论MCP从组织的真实数据源读取和写入数据。
由于代理通过本体论MCP执行的每个操作都会记录在本体论中，这种模式为无监督操作提供了可审计性。您可以使用任何代理框架（如Anthropic SDK或Google ADK）构建无头工作流，这些工作流可以响应本体论变化或按计划运行。