Realtime AI chat with AI Elements
This example shows how to build a realtime, multiplayer AI chat with Liveblocks Feeds, Next.js, and AI Elements.
Each chat is a feed inside a Liveblocks room, so messages sync instantly to
everyone connected, complete with shared presence (avatar stack and a live "AI
is thinking…" status). The AI reply is generated on the server and streamed back
into the feed with @liveblocks/node (createFeedMessage + updateFeedMessage),
then rendered live for all users through the useFeedMessages hook — including
reasoning, chain of thought, tool calls, sources, and token usage via AI
Elements.
Getting started
Run the following command to try this example locally:
This will download the example and ask permission to open your browser, enabling you to automatically get your API key from your liveblocks.io account.
Manual setup
Read more
Alternatively, you can set up your project manually:
- Install all dependencies with
npm install - Create an account on liveblocks.io
- Copy your secret key from the dashboard
- Create an
.env.localfile and add your secret key as theLIVEBLOCKS_SECRET_KEYenvironment variable - (Optional) Add an
AI_GATEWAY_API_KEYfrom the Vercel AI Gateway to get real model responses. Without it, the example uses a built-in mock assistant so it still runs end to end. - Run
npm run devand go to http://localhost:3000
To see the realtime sync, open the page in two browser tabs and send a message from one — it appears instantly in both, along with the AI's reply.
Deploy on Vercel
Read more
To both deploy on Vercel, and run the example locally, use the following command:
This will download the example and ask permission to open your browser, enabling you to deploy to Vercel.
Develop on CodeSandbox
Read more
After forking
this example
on CodeSandbox, create the LIVEBLOCKS_SECRET_KEY environment variable as a
secret.
