Agent Demos

Turn agent prototypes into deployable apps

Pick a production-ready demo slice, hand the guide to Codex, and ship a working Next.js agent app in about an hour. Already have a LangGraph agent? Use this as the launch path to a real web product.

GitHub

Ready now

16

Roadmap

0

Ready demos follow stable AI SDK Recipes and docs sources while roadmap cards stay separate.

shadcn registry guide

Start a new AI agent project from this registry

The guide gives humans and coding agents one path: create a styled shadcn Next.js app, add Foundation Chat from the registry, run a real AI Gateway chat check, and prepare the same env for Vercel.

agent task brief

Read https://agent-demos.hsawana9.com/registry-guide first. Then create a new shadcn Next.js app, install @agent-demos/foundation-chat, configure AI_GATEWAY_API_KEY, verify one local chat message, and prepare the Vercel deployment env.

Recommend

Start here

4 highlighted demos

ReadyFoundation
Number One
Ultra Chatbot Agent
An application-shape chatbot port with visitor-owned URLs, model selection, Postgres persistence, and resumable streams as the base for the full vercel/chatbot capability set.
App-shape portFoundation

Source: Pinned vercel/ai-chatbot snapshot ported into this repo architecture

ultra-chatbot-agentOpen demo
ReadyGenerative UI
Number Two
Generative UI
A chat workspace where the model can use hosted web search when recency matters, then choose a comparison matrix or recommendation card as the primary UI output.
UI toolsGenerative UI

Source: AI SDK UI Generative User Interfaces guide

generative-uiOpen demo
ReadyLangGraph
Number Three
LangGraph Agent
A Next.js and AI Elements frontend wired to the official LangGraph thread-scoped Agent Server streaming API.
LangGraph bridgeLangGraph

Source: LangGraph Agent Server API plus AI SDK UI message streams

langgraph-agentOpen demo
ReadyFoundation
Number Four
OpenAI Agents SDK Demo
An official OpenAI Agents SDK backend bridged into the existing AI SDK UI workspace without rewriting the agent runtime into repo-local orchestration.
Official bridgeFoundation

Source: OpenAI Agents SDK guides and official AI SDK UI bridge example

openai-agents-sdk-demoOpen demo

Ready demos

Interactive now

16 of 16 demos

ReadyFoundation
Foundation Chat
A production-ready base chat wired to AI Gateway and AI SDK 6, built as the copyable starting point for future demos.
Base chatFoundation

Source: AI SDK 6 stable runtime contract

foundation-chatOpen demo
ReadyRAG
RAG Chatbot
Knowledge-base ingestion and retrieval over durable storage, following the stable AI SDK recipe with a productized workspace.
Knowledge baseRAG

Source: AI SDK 6 stable RAG recipe

rag-chatbotOpen demo
ReadyMultimodal
Multi-Modal Chatbot
Chat over user-provided images and PDFs in a single turn, preserving the official AI SDK multimodal guide with a productized workspace.
Mixed inputMultimodal

Source: AI SDK 6 stable multimodal guide

multimodal-chatbotOpen demo
ReadyStructured Output
Object Generation
Generate a structured object from text, images, and PDFs, then render that object directly inside the assistant message.
Structured objectStructured Output

Source: AI SDK 6 structured output and useObject docs

object-generationOpen demo
ReadyGenerative UI
Generative UI
A chat workspace where the model can use hosted web search when recency matters, then choose a comparison matrix or recommendation card as the primary UI output.
UI toolsGenerative UI

Source: AI SDK UI Generative User Interfaces guide

generative-uiOpen demo
ReadyTools
Memory & Persistence Agent
Persist chat threads, explicit memory-tool writes, and handoff compactions so an agent can resume work across sessions.
Memory loopTools

Source: AI SDK memory, persistence, and embeddings docs

customer-memory-agentOpen demo
ReadyFoundation
Persistent & Resume Agent
A URL-backed agent chat that persists messages in Postgres, isolates visitors with an HTTP-only cookie, and resumes live streams after refresh.
Persistent chatFoundation

Source: AI SDK UI message persistence and resume streams

persistent-agentOpen demo
ReadyFoundation
Streaming Chat Shell
A developer-facing chat runtime shell that demonstrates shared useChat state, feature-local request metadata, and a replayable SSE trace.
Streaming shellFoundation

Source: AI SDK 6 chat and streaming runtime recipes

streaming-chat-shellOpen demo
ReadyLoop Agent
Loop Agent
A support triage agent that shows parallel context lookup, dependent SLA checks, human approval, visible tool state, and bounded loop control.
Approval gateLoop Agent

Source: AI SDK 6 stable tool-calling, loop-control, and HITL recipes

loop-agentOpen demo
ReadyLangGraph
LangGraph Agent
A Next.js and AI Elements frontend wired to the official LangGraph thread-scoped Agent Server streaming API.
LangGraph bridgeLangGraph

Source: LangGraph Agent Server API plus AI SDK UI message streams

langgraph-agentOpen demo
ReadySkills
Skills Builder Agent
A sandbox-backed ToolLoopAgent that loads repo-local skills on demand, pressures a rough idea into durable context, and drafts reusable SKILL.md artifacts.
Skill folderSkills

Source: AI SDK Add Skills to Your Agent guide

skills-agentOpen demo
ReadySandbox
Sandbox Workspace Agent
A persistent Vercel Sandbox workspace that lets the agent generate frontend files, run commands, and publish a live preview.
Live previewSandbox

Source: Sandbox workspace and live preview batch

sandbox-agentOpen demo
ReadyMCP
MCP Runtime Doctor Agent
An AI SDK ToolLoopAgent that discovers MCP tools from project docs and a local Next.js runtime, then answers repo and runtime questions through namespaced tool calls.
Connector hubMCP

Source: AI SDK MCP tools and Next.js dev MCP guide

mcp-agentOpen demo
ReadyFoundation
OpenAI Agents SDK Demo
An official OpenAI Agents SDK backend bridged into the existing AI SDK UI workspace without rewriting the agent runtime into repo-local orchestration.
Official bridgeFoundation

Source: OpenAI Agents SDK guides and official AI SDK UI bridge example

openai-agents-sdk-demoOpen demo
ReadyTools
Trace and Eval Agent
A live research agent that searches the web through AI Gateway, then scores the current conversation with source, answer-shape, and expected-path checks.
Trace + evalTools

Source: AI SDK 6 Gateway search, telemetry, and testing recipes

trace-eval-agentOpen demo
ReadyFoundation
Ultra Chatbot Agent
An application-shape chatbot port with visitor-owned URLs, model selection, Postgres persistence, and resumable streams as the base for the full vercel/chatbot capability set.
App-shape portFoundation

Source: Pinned vercel/ai-chatbot snapshot ported into this repo architecture

ultra-chatbot-agentOpen demo