Vercel's Chat SDK: One Codebase to Rule All Chat Platforms

Last month (Feb 23, 2026), Vercel quietly open-sourced something that should make every dev building chat features breathe a huge sigh of relief: the Chat SDK (npm i chat).

If you’ve ever tried to ship the same AI assistant on Slack and Discord and Teams, you know the drill. It’s less “build cool stuff” and more “translate between five angry APIs while questioning your life choices.”

Well, friends… the nightmare is over. Write your bot logic once. Deploy everywhere. And when you pair it with Vercel’s AI SDK? Pure magic – real-time streaming responses that actually look native on each platform.

No more copy-paste hell. No more “works on my Slack but not on Teams.” Just one happy codebase and happy users.

The Old Way: Platform Fragmentation Hell (We’ve All Been There)

You want an AI support bot in Slack for the team, a community helper in Discord, and something enterprise-friendly in Teams?

Classic move:

• Learn Slack’s threading quirks
• Figure out Discord’s slash commands
• Pray Microsoft Teams doesn’t break your cards
• Maintain three (or more) codebases
• Spend your weekends debugging emoji reactions

It’s like trying to teach the same joke in five different languages while the audience keeps changing the rules. Exhausting.

Enter Chat SDK: One Codebase, Zero Drama

The Chat SDK is a unified TypeScript library that turns this mess into a beautiful, event-driven dream.

You define your bot once, plug in adapters for each platform, and boom – consistent behavior everywhere. Mentions, reactions, buttons, modals, file uploads… it all just works.

Here’s how dead simple setup looks:

import { Chat } from "chat";
import { createSlackAdapter } from "@chat-adapter/slack";
// add discord, teams, github, linear, etc.

export const bot = new Chat({
  userName: "my-ai-sidekick",
  adapters: {
    slack: createSlackAdapter(),
    // one line per platform
  },
  state: createRedisState(), // or whatever you like
});

Add a new platform later? Just drop in another adapter. Your core logic doesn’t change. I felt my blood pressure drop just typing that.

Why It’s a Match Made in Heaven with the AI SDK

This is where things get really good.

Chat SDK’s post() method happily accepts AI SDK text streams. That means you can spin up a full ToolLoopAgent (or whatever you’re cooking with Claude/OpenAI/etc.) and stream responses in real time – with native formatting on every platform.

Check this beauty:

import { ToolLoopAgent } from "ai";

const agent = new ToolLoopAgent({
  model: "anthropic/claude-4.6-sonnet",
  instructions: "You are a helpful assistant.",
});

bot.onNewMention(async (thread, message) => {
  const result = await agent.stream({ prompt: message.text });
  await thread.post(result.textStream); // streams beautifully everywhere
});

One AI brain. Every chat app. Streaming. No extra glue code.

It’s the kind of integration that makes you whisper “thank you” to your monitor.

Why Chat SDK Actually Matters for Your Apps

This isn’t just another SDK to ignore in your package.json. Here’s the real value:

1. Time you’ll actually get back – No more duplicating features across platforms. Spend it on actual product instead of API wrangling.
2. Consistent AI experience – Your users get the same smart replies whether they’re in Slack during work or Discord after hours.
3. Reach without the rewrite – Internal tools, community bots, customer support… suddenly every chat platform becomes fair game.
4. Future-proofing that doesn’t suck – New platform pops up? Adapter. Done. (They’ve already added WhatsApp, tables, streaming markdown… this thing moves fast.)
5. Type-safe JSX cards – Build rich UIs once and they render natively everywhere. Buttons, tables, the works.

In a world where AI is eating everything, being able to drop intelligent chat into the tools people already live in? That’s not a nice-to-have. That’s table stakes.

The Humorous Silver Lining

Remember the good old days when you’d push a “small” update and suddenly your Discord bot started replying in binary while Slack went silent?

Those days are dead. Now you can break things in one place and fix them in one place. (Still test thoroughly though – we’re not animals.)

My favourite part? The team can finally stop asking “Does this work on Teams yet?” and start asking “What cool AI thing should we build next?”

Final Thought

If your app involves any kind of conversation (and let’s be honest, most do now), Vercel’s Chat SDK + AI SDK combo is the upgrade you didn’t know you needed.

Go npm i chat right now. Your future on-call self (and your blood pressure) will thank you.

Check it out: https://chat-sdk.dev/

Until next time – may your bots be consistent, your streaming smooth, and your platform APIs finally shut up and behave.

— Richard

P.S. The official changelog is a fun read if you want the full nerdy details: https://vercel.com/changelog/chat-sdk 😏