Uploaded April 2026 | Updated September 2026, 1 week ago
Build your agent logic once and deliver it to every platform with Chat SDK.
The Chat SDK is an open-source TypeScript library developed by Vercel to help developers bring AI agents to multiple messaging platforms (like Slack, Microsoft Teams, WhatsApp, etc.) using a single codebase. It focuses on the communication layer, allowing you to build rich, interactive chat interfaces using familiar tools like JSX and Markdown.
Video Chapters
00:59 | Malte (CTO of Vercel) explains how the need for a unified way to build Slack bots and other platform agents led to the SDK's creation during the 2025 holiday break.
03:24 | Discussion on how the SDK is a pure TypeScript library, not locked to Vercel infrastructure, and can be hosted anywhere that supports JavaScript.
04:41 | How the Teams team at Microsoft took ownership of the adapter, ensuring it uses the latest APIs with zero code changes for developers.
06:55 | The benefits of using JSX to render interactive elements (like buttons and menus) across different platforms through an intermediate UI representation.
08:53 | A technical breakdown of how JSX acts as "syntactic sugar" for function calls and how it integrates with existing React apps without breaking them.
10:36 | Managing differences in native rendering and threading (e.g., platforms with threads like Slack vs. context-free ones like WhatsApp).
13:23 | Clarifying that Chat SDK is the platform layer (bringing the agent to users), while AI SDK is the intelligence layer (building the agent itself).
15:59 | Using the SDK to listen to conversations and map them to external APIs like Jira or Linear.
17:28 | Categorizing adapters into "Official Vercel" (100M+ users), "Vendor Official," and "Community" adapters.
20:02 | A guide for developers to get started by looking at existing implementations and inheriting from the Base Class.
21:12 | Using JSX buttons to request user approval before an agent performs a sensitive action.
24:13 | Leveraging durable compute to allow agents to "await" a human interaction or webhook for long periods (even weeks).
25:44 | Call for community contributions to expand support for platforms like RCS, WeChat, and more.
Build your agent logic once and deliver it to every platform with Chat SDK.
The Chat SDK is an open-source TypeScript library developed by Vercel to help developers bring AI agents to multiple messaging platforms (like Slack, Microsoft Teams, WhatsApp, etc.) using a single codebase. It focuses on the communication layer, allowing you to build rich, interactive chat interfaces using familiar tools like JSX and Markdown.
Video Chapters
00:59 | Malte (CTO of Vercel) explains how the need for a unified way to build Slack bots and other platform agents led to the SDK's creation during the 2025 holiday break.
03:24 | Discussion on how the SDK is a pure TypeScript library, not locked to Vercel infrastructure, and can be hosted anywhere that supports JavaScript.
04:41 | How the Teams team at Microsoft took ownership of the adapter, ensuring it uses the latest APIs with zero code changes for developers.
06:55 | The benefits of using JSX to render interactive elements (like buttons and menus) across different platforms through an intermediate UI representation.
08:53 | A technical breakdown of how JSX acts as "syntactic sugar" for function calls and how it integrates with existing React apps without breaking them.
10:36 | Managing differences in native rendering and threading (e.g., platforms with threads like Slack vs. context-free ones like WhatsApp).
13:23 | Clarifying that Chat SDK is the platform layer (bringing the agent to users), while AI SDK is the intelligence layer (building the agent itself).
15:59 | Using the SDK to listen to conversations and map them to external APIs like Jira or Linear.
17:28 | Categorizing adapters into "Official Vercel" (100M+ users), "Vendor Official," and "Community" adapters.
20:02 | A guide for developers to get started by looking at existing implementations and inheriting from the Base Class.
21:12 | Using JSX buttons to request user approval before an agent performs a sensitive action.
24:13 | Leveraging durable compute to allow agents to "await" a human interaction or webhook for long periods (even weeks).
25:44 | Call for community contributions to expand support for platforms like RCS, WeChat, and more.
