Vercel Eve从入门到精通：构建部署企业级AI智能体 – CG群

本课程面向JS/TS开发者，教授用Vercel Eve框架构建部署企业级AI代理，涵盖文件系统优先架构、强类型工具、技能注入、MCP实时检索及人工审批闸，带你掌握下一代AI工程技能。

原始标题：Vercel Eve Beginner to Pro: Build & Deploy Agentic AI Agents

Vercel Eve Beginner to Pro: Build & Deploy Agentic AI Agents

本课程是一门面向前端与全栈开发人员的AI代理（Agentic AI）前沿实战课，核心教授如何利用 Vercel 最新的开源框架 Vercel Eve 构建并部署具备企业级生产力的自主 AI 智能体。课程采用创新的“文件系统优先（Filesystem-first）”架构理念，摆脱了传统聊天机器人的玩具属性，通过带领学员从零完整开发一套对接 Slack、用于真实房地产中介的智能多通道应用，深度解锁 AI 代理的核心构建块：包括使用强类型工具读写真实数据库、通过“技能（Skills）”注入行业知识、利用模型上下文协议（MCP，如 Exa MCP）进行全网实时检索，以及通过 Cron 实现无人值守的自动化定时任务。同时，课程重点攻克了安全落地难题，传授如何设计“人工审批闸（Human Approval Gates）”以确保高风险操作的安全可控，旨在帮助具备 JS/TS 基础的工程师抢先掌握下一代 AI 软件工程的黄金技能。

Use Vercel Eve AI to build and deploy agentic AI agents. With tools, skills, MCP, schedules, human approvals, + more!

What you’ll learn
⚡ Build AI agents the filesystem-first way with Vercel Eve, where a folder of files is your agent
⚡ Connect an agent to a real database using typed tools that create and read records
⚡ Give your agent domain knowledge and procedures with on-demand skills
⚡ Integrate external services over MCP (Model Context Protocol)
⚡ Put risky actions behind human approval gates that appear as buttons your team taps
⚡ Schedule autonomous tasks with cron – work that runs every day with nobody present

Requirements
❗ A free Vercel account (we’ll set up everything else together)
❗ No prior experience with AI agents or Vercel Eve required – we start from zero

DescriptionAI agents are everywhere, but most tutorials stop at a toy chatbot that can’t do anything real. This course is different. You’ll learn Vercel Eve, Vercel’s open-source framework for building production AI agents, by building one complete, genuinely useful project, from an empty folder all the way to a deployed, multi-channel application.

The project you’ll build
Your project is a real-world AI assistant for areal estate sales agency. The kind of software a business would actually pay for. From Slack, real-estate agents talk to your Eve agent in plain language to

✨ add and update property listings (using custom tools)

✨ search inventory against a buyer’s exact requirements

✨ mark properties as sold

✨ book viewings, and receive a daily “your viewings today” digest delivered automatically (using scheduled tasks)

✨ see sales performance as charts generated on the fly

✨ and write polished, accurate propertyarea descriptions by researching the live web (using Exa MCP)

It’s all driven by a single database, andevery feature is chosen to teach a specific Eve building block, so by the end, the patterns are yours to reuse on any agent, in any domain. The real estate app is just the vehicle; the skills transfer everywhere.

Why Vercel Eve, and why now
Eve is new, which means the developers who learn it now are early. It’s built on Vercel, so you get serverless deployment, scheduled jobs, sandboxed compute, managed authentication, and model routing out of the box. And its filesystem-first design means adding a capability is as simple as adding a file.

What makes this course different
This isn’t a string of disconnected demos. It’sone coherent project that grows module by module, the way you’d actually build software, and it doesn’t dodge the hard parts.

The examples are deliberately real and useful, no throwaway projects, but they stay simple enough that you can follow along and build every piece yourself.

By the end you won’t just understand Vercel Eve, you’ll havebuilt and shipped a real, multi-channel AI agent, and you’ll know how to design, secure, schedule, and deploy your own.

Who this course is for
⭐ Developers comfortable with JavaScript or TypeScript who want to build real, deployable AI agents
⭐ Anyone who’s called an LLM API or used ChatGPT and wants to learn how to turn a model into actual software
⭐ Engineers who want an early, in-demand skill in the fast-moving AI agent space

隐藏内容

此处内容需要权限查看