ZERO TO AGENT

Mohali
Track Guide

Your step-by-step companion for building and deploying an AI agent. Pick a track, follow the steps, ship something real.

✦ Find Your Track

⚙️We recommend: Workflow (WDK)

Based on your answers, this track is the best fit. But you can choose any track below!

✦ Pre-Event Setup

Do this before you arrive — or we'll help you at the start. Tap to check off.

Everyone
Create a free Vercel accountvercel.com
Create a v0 accountv0.dev
Bring your laptop + charger
Have a Google or GitHub account ready
Bookmark the Vercel docsvercel.com/docs
Tracks 1 & 3 Only (Developers)
Install Node.js (v18+)nodejs.org
Install VS Code or your preferred editorcode.visualstudio.com
Install Gitgit-scm.com
Test: run npx create-next-app@latest
Verify: node -v && npm -v

✦ The Tracks

🚀v0 + MCPsBeginner
★★★★★

Build with words, not code

▾
⚙️Workflow (WDK)IntermediateRECOMMENDED
★★★★★

Durable agents that survive anything

▾

Build long-running, durable async agents with the Workflow Development Kit. Your agents survive crashes, resume after deploys, and can pause for minutes or months.

Ideal for: Developers comfortable with JavaScript/TypeScript and Next.js basics

Step-by-Step Guide

1
Scaffold a Next.js App▾

Run: npx create-next-app@latest --use-npm

2
Add WDK▾

Run: npm install @vercel/workflow

3
Configure Next.js▾

Wrap your Next.js config with the workflow configuration to enable the runtime.

4
Create Workflow Functions▾

Use 'use workflow' directive and define steps with 'use step'. Each step is independently retriable.

5
Add AI with StreamableAgent▾

Import from @vercel/ai/react and pair with the AI SDK for AI-powered workflow steps.

6
Get a Gateway API Key▾

Go to the Vercel AI Gateway dashboard and generate an API key for your project.

7
Deploy to Vercel▾

Push to GitHub or deploy via CLI. Vercel auto-provisions the workflow infrastructure.

⭐ Pro Tips for This Track

⭐ Use the AI Gateway
Don't juggle API keys. The AI Gateway gives you a single endpoint for every major model provider with automatic fallbacks and rate limit handling.
⭐ Deploy Early, Test on Vercel
WDK persistence only works on Vercel — local dev runs workflows but without durability. Push to Vercel as soon as your first step works so you can test the real behavior.
⭐ Use AI SDK DevTools
AI SDK 6 has built-in DevTools for debugging multi-step agent flows. You get full visibility into LLM calls, tool use, and execution trajectories.
⭐ Feed Docs to Your Coding Agent
If you're using a coding agent (Claude Code, Cursor, etc.), feed it the WDK docs from caseswflow.dev and the AI SDK docs from sdk.vercel.ai/llms.txt for accurate code generation.
⭐ Add the Vercel Plugin
If you're using a coding agent, add the Vercel plugin for best-practice guidance. Run: npx vercel-ai-sdk for ChatSDK or equivalent for WDK skills.

⚠️ Common Issues

⚠️ "My workflow isn't persisting"+
⚠️ "Steps aren't retrying on failure"+
⚠️ "Config errors on build"+
⚠️ "Can't access AI Gateway"+
💻ChatSDK AgentsAdvanced
★★★★★

Write once, deploy everywhere

▾

Build agents using Vercel AI SDK + AI Gateway + ChatSDK that interface across Slack, Discord, Teams, GitHub, and more. Write your bot logic once, then deploy to every platform via swappable adapters.

Ideal for: Experienced developers comfortable with SDKs, npm packages, and terminal workflows

Step-by-Step Guide

1
Install Core Packages▾

Run: npm install @chat-sdk/core @chat-sdk/adapter-slack @chat-sdk/adapter-discord (pick your platforms)

2
Pick a State Adapter▾

Choose production storage (Redis, Postgres) or state-memory for local dev.

3
Create a Shell Instance▾

Initialize your Shell with your chosen adapters and configuration.

4
Wire Up Handlers▾

Set up bot.onNewConversation, bot.onNewMessage, bot.onReaction, and other event handlers.

5
Add AI▾

Use 'model.config' — the Shell accepts AI SDK text streams natively. No extra wiring needed.

6
Use the AI Gateway▾

Configure zero-config access to any model provider (OpenAI, Anthropic, Google, etc.) through a single endpoint.

7
Deploy to Vercel▾

Push to Vercel — adapters auto-detect credentials from environment variables.

⭐ Pro Tips for This Track

⭐ Use the AI Gateway
One endpoint for every model provider. The Gateway handles rate limits, automatic fallbacks, and provider switching — no per-provider API key management.
⭐ Start With One Adapter, Add More Later
Don't try to ship to Slack, Discord, and Teams all at once. Get one adapter working, deploy it, then add the next. Your bot logic stays identical across platforms.
⭐ Use AI SDK DevTools
AI SDK 6 has built-in DevTools for debugging multi-step agent flows. Full visibility into LLM calls, tool use, and trajectories. Essential for ChatSDK debugging.
⭐ Deploy Early, Iterate Fast
Push to Vercel after your first onNewMessage handler works. Every git push creates a preview deployment. Adapters auto-detect credentials from env vars on Vercel.
⭐ Feed Docs to Your Coding Agent
Using Claude Code or Cursor? Feed it the ChatSDK docs from chat-sdk.dev and AI SDK docs from sdk.vercel.ai/llms.txt for accurate code generation.

⚠️ Common Issues

⚠️ "Adapter won't connect to Slack"+
⚠️ "Bot responds but no streaming"+
⚠️ "State isn't persisting between messages"+
⚠️ "Multiple adapters conflicting"+
🏆

Judging Criteria

Agent Usefulness
Does it solve a real problem?
Technical Execution
Built with v0, deployed on Vercel
Creativity
Unique idea or novel approach
$6,000+ in prizes • Up to $3,000 in v0 credits for 1st
BUILD IT. SHIP IT. LET'S GO.