https://downcity.ai/en/blog/ Long-form explanations of the agent runtime layer and agent productization. en https://downcity.ai/en/blog/agent-harness-vs-mcp/ https://downcity.ai/en/blog/agent-harness-vs-mcp/ MCP connects an agent to tools. A harness runs the agent. Here is why they get compared, where each one's boundary actually sits, and how they compose in practice. Fri, 18 Sep 2026 00:00:00 GMT https://downcity.ai/en/blog/the-anatomy-of-an-agent-harness/ https://downcity.ai/en/blog/the-anatomy-of-an-agent-harness/ A component-by-component tour of an agent harness. One request traced end to end, ten subsystems and what each one owns, and a diagnostic table for when something breaks in production. Fri, 18 Sep 2026 00:00:00 GMT https://downcity.ai/en/blog/what-is-an-agent-harness/ https://downcity.ai/en/blog/what-is-an-agent-harness/ An agent harness is the runtime layer that turns a language model into a working agent. Here is what it owns, how it differs from a framework or MCP, and why production raises the bar. Fri, 18 Sep 2026 00:00:00 GMT https://downcity.ai/zh/blog/agent-harness-vs-mcp/ https://downcity.ai/zh/blog/agent-harness-vs-mcp/ MCP 负责把 Agent 连到工具，Harness 负责运行 Agent。本文说明两者为什么常被放在一起比较、各自的边界在哪里，以及实际工程中它们如何组合。 Fri, 18 Sep 2026 00:00:00 GMT https://downcity.ai/zh/blog/how-to-choose-an-agent-framework/ https://downcity.ai/zh/blog/how-to-choose-an-agent-framework/ 选智能体框架之前，先分清编排库、低代码平台和 Agent 运行时是三类不同的东西。本文给出四个判断维度、一份选型清单，以及最容易踩的三个误区。 Fri, 18 Sep 2026 00:00:00 GMT https://downcity.ai/zh/blog/the-anatomy-of-an-agent-harness/ https://downcity.ai/zh/blog/the-anatomy-of-an-agent-harness/ 逐个拆解 Agent Harness 的组件：从一条请求的完整链路讲起，说明十个子系统各自负责什么，并给出一张「出问题时该查哪个器官」的排查表。 Fri, 18 Sep 2026 00:00:00 GMT https://downcity.ai/zh/blog/what-is-an-agent-harness/ https://downcity.ai/zh/blog/what-is-an-agent-harness/ Agent Harness 是把大模型变成可用 Agent 的运行时层。本文说明它到底负责什么、和 Agent 框架及 MCP 的区别，以及为什么进入生产环境后要求会突然变高。 Fri, 18 Sep 2026 00:00:00 GMT