Agent Z

Agent Z powers a community-built Discord server structured like a real hackathon hub—welcome and onboarding, announcements, general chat, show-and-tell, help / debugging, and building-in-public—so newcomers know where to go and the bot can guide them in-channel. Built on Vercel Chat SDK with PostgreSQL, optional Gateway for mentions/replies, tier-gated Discord tools over MCP, and a small admin surface. You’re invited to join our live server to see the flows, slash commands, and bot behavior exactly as members experience them.

Lio OS

Create your own custom OS based on your workflow using AI

Synapse

Synapse is a personal knowledge orchestrator that builds a real model of each student—how they learn, what they know, what they struggle with—and uses that model to make decisions in every interaction. Instead of summarizing content, it conducts active validation sessions based on the user's own material: open-ended questions, exercises, and the Socratic method. It detects when something isn't working and adjusts its approach in real time. It remembers progress across sessions and schedules reviews using FSRS-6, the most accurate spaced repetition algorithm available today. Built on LangGraph for stateful per-user memory, Vercel AI Gateway for LLM routing, and Neon for vector persistence—all within an architecture designed to scale without changing the tech stack.

Spotlight JLM

Spotlight JLM reimagines event discovery in Jerusalem by using AI to match people with experiences based on mood, not just time and location. It unifies diverse communities across West and East Jerusalem - into one seamless, inclusive platform. It holds the information about all types of events in Jerusalem - from official to underground, cause the AI Agent knows to search through all the platforms. It can accurately detect mood of who asking.

Aireaviso

Aire

DeployDoctor

An AI-powered Vercel deployment log diagnosis tool that turns noisy failed build logs into clear, actionable repair reports. Developers paste a deployment log, get the likely root cause, evidence, confidence level, files to check, commands to run, and next fix steps, then optionally share a sanitized public diagnosis page with teammates. It uses server-side Cerebras inference with mock fallback and is designed around privacy: raw logs are not persisted, secrets are redacted, and shared pages store only sanitized diagnosis results.

Salt, Salt - The game

Get all the salt into the salt jar and make your own levels with the level mcp

Agentic SEO Orchestrator

An automated, multi-agent platform designed to execute technical SEO optimizations and structural UI improvements entirely through team chat. Leveraging the Vercel AI SDK and v0 MCP, it acts as an asynchronous SEO engine within Discord. The system researches SEO gaps, generates fully styled Next.js React components via v0 to address them, and facilitates stakeholder consensus directly in the chat thread. Once approved, the orchestrator commits the finalized UI components to the connected GitHub repository, triggering an automatic Vercel deployment without breaking existing CI/CD pipelines.

Riad

Riad gives every employee a calm, voice-first companion that body-doubles them through enrollment and stays present for the full plan year. Built with the Vercel's WDK, AI Gateway server, and a custom benefits knowledge MCP.

SideBet

Side Bet is a private social prediction market for coursework. Friends bet cosmetic dollars on assignment outcomes, while an AI bookmaker prices markets from submission history and flags early signs of academic disengagement. Built with Next.js, Supabase, LMSR market math, realtime updates, and AI-powered odds generation.

AI Tax Assistant

An AI-powered web application built with v0 that helps users understand and calculate their income tax in Sri Lanka. The assistant allows users to ask natural language questions like “How much tax do I pay for 150,000 LKR?” and provides accurate answers by calling backend tools through an MCP server. It uses the Vercel AI SDK for real-time responses and integrates custom MCP tools such as tax calculation and tax rule retrieval, ensuring reliable, tool-driven outputs instead of guesswork. The app features a simple chat interface and demonstrates how AI agents can interact with structured services to deliver practical financial insights.

ServiceHub

I built a small directory that ranks nearby plumbers, electricians, tutors, cleaners, or repair workers using reviews, availability, and price estimates.

ShipGuard Agent

ShipGuard Agent is an AI DevOps recovery agent for Vercel teams. Each developer signs in with their own Vercel account, connects their real projects, and gets an incident command center for failed deployments. When a deployment fails, ShipGuard fetches Vercel deployment data and logs, verifies affected pages with Firecrawl, analyzes root cause with Gemini, generates a safe recovery plan, waits for human approval, and then verifies recovery before marking the incident resolved. The goal is to reduce the time between “my deployment failed” and “production is healthy again” with a durable, auditable workflow instead of manual log hunting.

sniff

Sniff is an AI-powered GitHub PR code reviewer built with Next.js , you paste a GitHub PR URL, pick a reviewer persona, and an LLM (via Groq/OpenAI .etc) performs a deep, structured code review. It can fetch file contents and list directories directly from the repo to ground its analysis — no hallucination, tool-first approach. Three personas available: Strict Senior — blunt, zero-tolerance, blocks on missing tests Friendly Mentor — Socratic, explains the "why", teaches as it reviews Code Troll — technically precise, wrapped in surgical humor The output is a structured review in Spanish with sections for summary, issues (by severity), positives, and a final verdict.

Scoutfolio

ScoutFolio is an AI portfolio builder for students: you connect GitHub's MCP (more incoming), a resume PDF, or URLs, and server-side agents extract signal, pick what is worth showing, and turn it into recruiter-style copy plus a ranked inventory. From that inventory, another agent uses Next.js scaffolding so you get a preview with an active agent chat to make changes, code view, and zip export of a deployable personal site on the Vercel-oriented stack (Next.js, auth, Blob, Supabase, rate limits). The product story is “from no portfolio to something you can ship,” with streaming agent UI for demos.

Offline Talking AI JARVIS!! <<((((

This is a ready-to-distribute offline Talking AI assistant package! Single download, complete app, no configuration needed. ✅ **Complete offline operation** - No internet after setup✅ **Cross-platform** - Windows, macOS, Linux✅ **Voice interaction** - Speak to AI, hear responses✅ **Multiple LLM support** - Ollama, LM Studio, LLaMA.cpp✅ **Easy distribution** - Professional installers✅ **User-friendly setup** - Guided wizard✅ **Production-ready** - Fully tested and documented✅ **Privacy-first** - All local processing, no data collection

PropInsight

Finding your future home is messier than it should be. Comparing multiple properties across prices, neighborhoods, schools, square footage, and a dozen other factors that shape daily life can quickly become exhausting. The stakes are certainly too high to rush it or get wrong. PropInsight was built to ease that burden. Whether you're browsing condos downtown, houses in the suburbs, or apartments across multiple cities, we pull it all together for you. No more juggling tabs, no more messy spreadsheets, no more second-guessing what you might have missed. Powered by Zillow data, PropInsight delivers a clean, side-by-side comparison of every property you're considering: pricing, square footage, mortgage estimates, neighborhood insights, school ratings, walkability, and much more. Now you can spend less time digging and more time deciding. PropInsight: Buy the RIGHT home, not just the pretty one.

Fitness Score based on Cardio Health

This is a fitness score web app that takes in a user's key cardio health metrics and predicts their score and provides some actionable recommendations to improve their cardio health.

My Second Brain

Do you also struggle with saving way faster information than you can actually consume? Information that you promised yourself that you'll revisit and check out later but never found the time to do so ? I have thousands and thousands of X bookmarks, YouTube videos, articles, notes and more!! (that i've not included in this hackathon), that I accumulate almost daily — but hardly any time to go through them one by one. I now realize it's simply impossible ! The backlog keeps on growing too fast, and I constantly forgot what I’d saved. “I know I saved something about this…” became a daily frustration. I wanted to be able to search them easily ! So I built My Second Brain — a personal knowledge platform that finally makes sense of everything I save. It automatically: - Extracts clean previews and key entities from links, videos, bookmarks, and notes - Lets you search everything instantly with hybrid (keyword + semantic) search - Auto-generates beautiful, interconnected wiki pages (in the spirit of Karpathy’s LLM Wiki) so you actually understand and remember what you saved - Gives you a delightful keyboard-first triage inbox so you can quickly keep, archive, or drop items It’s my personal second brain. Built in one week with v0, Next.js 16, Neon + pgvector and Vercel Workflows.

Genesis Resume

Genesis Resume is an intelligent, AI-driven platform that transforms your static CV into perfectly optimized, job-specific versions in seconds. How it works: Build once — Generate a professional resume from scratch or import your existing CV. Tailor instantly — Paste any job vacancy URL and let AI analyze the requirements, then intelligently rewrite achievements, skills, and experiences to match the role. Scale your applications — Manage multiple tailored resumes and cover letters in one beautiful dashboard. Note: Import CV functionality and multi-variant AI generation are not implemented yet. The current version demonstrates the core AI tailoring flow using a manually created base resume. Built with modern web technologies and powered by advanced AI, Genesis Resume helps job seekers stand out by speaking the exact language recruiters and ATS systems are looking for. No more manual tweaking for every application — just faster, smarter, and more effective job hunting. Live demo: https://v0-cv-dashboard-application.vercel.app/

DJ Sally

escription DJ Sally is a Vercel Workflow-powered DJ control and readiness agent for live performers. It connects a Vercel-hosted Next.js dashboard with two matching HID hardware controllers, mapping physical knobs and buttons to deck controls and Android/Bluetooth media volume.

Patch

Patch: The Self-Healing Code Agent Patch is an AI-powered autonomous agent that monitors GitHub repositories and resolves bugs without human intervention. The moment a new issue is opened on a connected repository, Patch begins a fully orchestrated multi-step investigation. It pulls the relevant source files via GitHub MCP, analyzes the stack trace and error context using Anthropic Claude, generates a targeted patch, validates it inside a Vercel Sandbox, and opens a pull request with a full explanation. If the fix cannot be confirmed, Patch posts a structured diagnostic comment so the engineer knows exactly where to look. The problem it solves Software teams lose hundreds of hours every month responding to the same class of bugs: null pointer exceptions, type mismatches, broken imports, and failed API calls. These issues are tedious to fix manually but too nuanced for simple automation. Patch closes that gap entirely. What makes it technically distinct Patch runs on Vercel's Workflow SDK as its orchestration backbone. Unlike a standard chatbot or single-prompt agent, Patch runs as a durable multi-step workflow that persists state across long-running operations. If a sandbox test takes five minutes, the workflow does not time out or lose context. It waits, resumes, and delivers a result. This makes Patch reliable enough for production, not just demos. The only thing a user provides is a GitHub Personal Access Token with repo and issues permissions. That token is used solely to read code and open pull requests on the connected repository. All AI processing runs server-side using secure environment variables. Nothing sensitive is ever stored in a database or logged. Key technologies Vercel Workflow SDK for durable orchestration. Anthropic Claude via AI SDK for code analysis and patch generation. GitHub MCP for code fetching and pull request creation. Vercel Sandbox for isolated test execution. Next.js 14 for the dashboard and API routes. Deployed on Vercel. Why it stands out Every other AI tool in this space is a chat interface. You describe the bug, it tells you what might be wrong. Patch closes the loop entirely. It detects, investigates, fixes, validates, and ships autonomously. That is the difference between an AI assistant and an AI agent.

Anti-FOMO Agent

Anti-FOMO Agent turns X demo posts into evidence-grounded PRDs and build briefs. Send Pepe a post on Telegram or paste one on the web, and it imports the media, runs the analysis, and returns a practical Markdown handoff for Codex, Claude Code, or v0. Built with Vercel Workflow, AI Gateway, Blob, Redis, AI SDK, and a Telegram chat interface.

Cold Reach

What it does Cold Reach turns cold outreach into a swipe deck. Filter — pick company stage, role, location, and industry. Generate — an AI agent searches Apollo for verified contacts, then drafts 2–3 personalized cold emails per person using Claude. Swipe — right sends the email via your real Gmail, left sends it to a queue for later editing. From "I need to reach out to fintech founders in SF" to a real email landing in their inbox — under a minute, no copy-paste, no tab juggling.