What will you build today?
Describe it.Ship the whole app.
Build full-stack apps through conversation — 54 starters across web, desktop, mobile and ten languages,18 of them ready-made AI SaaS dashboards, each with a real terminal and a live preview. Powered by 24 AI providers and MCP tools from Composio & 45+ presets.
Stack, design, model and MCP tools are set on the next step — everything you type here comes with you.
For the best experience, have ready: Anthropic API key · Composio API key · Firecrawl API key · a Supabase organization
From the community
Start from something that already works
24 integrated AI providers — plus connect your own custom models
- Claude
- GPT-5
- Gemini
- Mistral
- Groq
- DeepSeek
- OpenRouter
- Perplexity
- xAI Grok
- Cohere
- Fireworks
- Qwen
- NVIDIA
- Cerebras
- Moonshot Kimi
- Together AI
- Azure OpenAI
- Ollama
- LM Studio
- Fal.ai
- Eden AI
- CometAPI
- ZenMux
- Vercel AI Gateway
- + Custom
Full-Stack Next.js Apps, Built by Conversation
Not a snippet generator. A real project — auth, database, payments, email, API routes — scaffolded, run and deployed while you watch.
Full Next.js project
Routes, server actions, schema and API — a real repo, not a file.
Real terminal
xterm shells plus a streaming dev-server log, per project.
Live preview
A running sandbox you can inspect, resize and share.
Push to GitHub
Repo created, code pushed, auto-deploy wired up.
Deploy to Vercel
Production deploys, env vars synced, custom domains.
Pick your stack — it scaffolds the wiring
Supabase Auth · Clerk · Auth.js · Better Auth · Firebase
Supabase Postgres · Drizzle · Prisma · Firestore
Stripe · Lemon Squeezy · Polar
shadcn/ui · Radix · Mantine · Material UI
TanStack Query · SWR · Zustand · React Hook Form + Zod
Vercel AI SDK · Anthropic · OpenAI · Resend · Storage
54 starters — web, desktop, mobile and ten languages
Each one comes from its framework's own CLI or from a NewAgentsHub base repo, then runs in a sandbox with a real terminal and a live preview. Pick a starter and you get a form built for it — a Django project is asked different questions than a Next.js one. The AI SaaS row is a working product on 18 frameworks: a landing page, Supabase sign-in, AI chat with saved history, your MCP tools and knowledge search, wired to the builder's AI gateway and ready to extend.
Next.js · Vite + React · Vite + Vue · Vite + Svelte · Vite + Solid · SvelteKit · Nuxt · Astro · Astro Blog · React Router · Angular
Bootstrap · Bootstrap SaaS Dashboard
NewAgentsHub Desktop · Electron + React · Electron + Vue · Electron + Svelte · Electron + Vanilla
Expo (iOS, Android, Web) · Expo Tabs · Expo Blank · Capacitor + Vite
FastAPI · Flask · Django · Streamlit
Rails · Sinatra
Laravel · Symfony
net/http service
Axum service
Spring Boot
ASP.NET Core
Phoenix
Deno service
React · SvelteKit · Nuxt · Astro · React Router · Angular · Expo · Capacitor · FastAPI · Django · Go · Rust · Deno · Rails · Laravel · ASP.NET Core · Spring Boot · Phoenix
Or bring code you already have
Import a project and the builder works out what it is — language, framework, package manager, dev command and port — then runs it. Every language above is detected, so a Rails or FastAPI repo arrives ready to preview rather than as a folder of files.
Any public repo — paste the URL.
gitlab.com or your own self-hosted instance.
A .zip of the project, straight from disk.
Pick a folder and edit it in place.
No URL in mind? Describe what you want instead and it searches open-source starters for you, then imports the one you pick.
15 coding agents in the terminal — not only ours
Every project gets a real shell, and these are already wired for it. Four ship in the sandbox image and are there the moment it boots; the other eleven install from the same menu in one command. They run on the provider keys you have already given us — you opt in once, and then nothing asks you to log in again.
Claude Code · Codex · opencode · pi
Gemini CLI · Grok · Qwen Code · Groq Code CLI
GitHub Copilot CLI · Cursor CLI · Factory CLI · Amp
Crush · Goose · Aider
Each one opens in the project's own directory beside a generated AGENTS.md that says what this app has connected — framework, MCP servers, which keys are in the environment — so it starts oriented instead of reading its way in. The shell is a tmux session: close the tab and the agent keeps working, come back and you reattach to the same one.
Five things that make it much better
The builder works without any of these. With them, it writes better code, wires a real database itself and researches as it goes.
- R1
Anthropic API key
The best coding model for the builder agent.
Get one at console.anthropic.com, then add it under Providers.
- R2
Composio API key
The MCP gateway the builder uses for Supabase and 100+ toolkits.
From composio.dev — paste it once in the MCP Connect dialog.
- R3
Firecrawl API key
Web research while building, design references and app thumbnails.
From firecrawl.dev — the builder asks for it the first time you turn on research.
- R4
A Supabase organization
Gives the apps you build a real database and sign-in.
Create one free at supabase.com/dashboard.
- R5
Supabase MCP through Composio
Lets the builder create and wire the database itself.
Connect Supabase in the Composio gateway, then authorize your organization.
Everything You Need in One Platform
From real-time chat to full AI apps — build, deploy, and scale with ease.
Multi-Provider Chat
Real-time streaming conversations with 24 AI providers. Switch models mid-chat, customize settings, and generate artifacts automatically.
MCP Server Integration
45+ preset MCP servers — just add your API key. Plus three gateways: Composio, Smithery and Docker. Or connect custom servers over HTTP, SSE, WebSocket or Stdio.
Next.js App Builder
FlagshipGenerate a complete Next.js project from chat — files, packages, migrations — then run it in a sandbox with a real terminal and deploy it to Vercel.
Custom Providers & Models
FlexibleConnect your own OpenAI-compatible endpoints, local models, or any custom API. Full control over authentication and parameters.
AI Agents
Create agents with custom system prompts, capabilities, and tool bindings. Assign roles like coder, researcher, analyst, or coordinator.
Agent Teams
Build multi-agent teams with 4 workflow types: Collaborative, Sequential, Democratic, or Hierarchical. Agents discuss and reach consensus.
Desktop & Mobile Apps
NewTurn any project into Windows, macOS, Linux, Android and iOS apps with one click. Installers build in your own GitHub Actions, and Android builds come back with screenshots from a real emulator.
App Projects
NewCombine providers, capabilities, agents, and MCP tools into deployable AI MCP Apps — or go further and generate a full Next.js project.
Knowledge Base (RAG)
Upload files, scrape URLs, process for retrieval. Supports PDF, DOCX, code files, and more. Automatic chunking for AI context.
Creations & Artifacts
Auto-generate code, documents, and data files from chat. View, edit, download, or export to PDF/DOCX. Full version history.
Build Custom AI Skills
Create, generate, or import capabilities. Give your agents superpowers.
Create Manually
Define custom capabilities with input parameters, output formats, and category assignments. Full control over every detail.
- Custom parameters
- Type validation
- Category organization
- Tag management
Generate with AI
Describe what you need in natural language. AI generates the complete capability definition automatically.
- Natural language input
- Auto-generated schema
- Instant testing
- One-click save
Import Skills
Import capabilities from Skill.md files or the community library. Share and reuse across projects.
- Skill.md format
- Community library
- Version control
- Easy sharing
24 Providers + Your Own
Connect to every major AI provider or bring your own models. Full parameter control with presets.
- Frontier labs
- Claude, GPT-5, Gemini, Mistral, Groq, DeepSeek, Perplexity, xAI Grok, Cohere, Fireworks, Qwen, NVIDIA, Cerebras, Moonshot
- Aggregators
- OpenRouter, Together AI, Vercel AI Gateway, Eden AI, CometAPI, ZenMux, Fal.ai
- Local & private
- Ollama, LM Studio, Azure OpenAI, and any OpenAI-compatible endpoint
- Full model settings
- Temperature, top-p, max tokens, penalties
- Auto-discovery
- Fetch available models straight from the provider APIs
- Secure storage
- Encrypted API keys, masked display
4 Gateways + 45 Presets
Composio joins Smithery and Docker as a first-class gateway — connect once, and every tool shows up in chat, in your agents and in the app builder.
- Composio
- One API key unlocks 1,000+ app toolkits (Gmail, Slack, GitHub, Notion…) with managed per-user auth and automatic token refresh
- Custom servers
- Any MCP server URL over HTTP, SSE or WebSocket, with OAuth or a bearer token
- Smithery
- Browse and connect servers from the Smithery registry
- Docker MCP Gateway
- Run MCP servers in containers seamlessly
- Every transport
- 45+ presets over HTTP, SSE, WebSocket and Stdio — plus OAuth, bearer tokens and custom auth
- Includes
- Playwright, FileSystem, Git, Postgres, Slack, Notion and more
From Zero to Shipped App in Minutes
Built for Every Team
From solo developers to enterprise teams — NewAgentsHub adapts to your workflow.
For Developers
Describe a full-stack app and watch it get built — real files, real terminal, real preview. Next.js, Django, Rails, Laravel, Spring Boot and more, without leaving the tab.
- Full Next.js projects with auth, database and payments
- Real xterm terminals and a live dev-server log
- GitHub push and Vercel deploy from the same window
- Multi-model testing across 24 providers, or your own
Simple, Transparent Pricing
Bring your own API key — you pay the model provider directly, we never mark it up. You pay us only for what runs on our side: sandbox hours and storage.
Free Trial — 7 days, no card required
All 24 AI providers on your own key · 3 hours of live sandbox · 500 MB storage
- All 24 AI providers on your own key — no markup
- 15 hours of live sandbox / month
- 1 GB storage
- 1 build at a time
- Agents, teams, MCP servers and deploys
- Community support
- Everything in Starter, plus:
- 50 hours of live sandbox / month
- 10 GB storage
- 2 builds at a time
- Email support (48h)
- Everything in Pro, plus:
- 120 hours of live sandbox / month
- 20 GB storage
- 3 builds at a time
- Priority support (24h)
- Everything in Pro, plus:
- 3 seats with shared workspaces and roles
- 40 hours of live sandbox / month, per seat
- 25 GB storage, per seat
- 5 builds at a time
- Priority support (24h)