Open AI Agents Hub

Open-source platform for building, browsing, and chatting with AI agents — both LLM chat agents and media generation agents (image, video, audio) — in a single web interface.

Instead of writing custom integration code for every model provider, Open AI Agents Hub gives you an agent library (your own agents, template agents, and featured agents), a chat/generation interface per agent, and a proxy backend so you can run the whole thing on your own infrastructure with your own API key.

🎨 Explore 50+ more open-source AI apps →

Why agents, not just chat

Most "chat with AI" tools stop at text. An agent here is a reusable, shareable unit that bundles:

• a system prompt / persona
• a target capability — chat, image generation, video generation, or audio generation
• its own profile page, slug, and conversation history

That means the same interface handles a customer-support chatbot agent and a "turn my photo into anime" image-generation agent, side by side in one library — you're not bolting media generation on as an afterthought.

Features

• Agent library — browse your own agents, ready-made template agents, and featured agents from one screen.
• Agent builder — create and edit agents visually (the client uses React Flow for the agent creation/edit canvas) with custom prompts, skills, and profile pages.
• Chat agents — multi-turn conversations per agent, with full conversation history stored per agent + conversation ID.
• Media generation agents — agents whose "reply" is a generated image, video, or audio clip instead of text.
• Suggested agents — an endpoint that recommends agents based on what you're trying to do.
• Like / profile pages — each agent has a public-style profile (/agents/{slug}/profile) so agents can be discovered and shared inside your instance.
• Bring your own key — the backend is a thin FastAPI proxy; you supply the credentials for whichever model/agent provider backs it, so there's no vendor lock-in baked into the UI.
• Self-hosted — Next.js frontend + FastAPI backend, run both locally or deploy anywhere that runs Node and Python.

Architecture

client/   Next.js 16 (React 19) app — agent library, agent builder (React Flow), chat + profile pages
server/   FastAPI backend — proxies agent CRUD, chat, and generation requests to your configured provider
packages/agents/  Shared React component library used by the client (agent cards, theming, create/edit UI)

The backend (server/app/routers/agent_proxy.py) exposes REST endpoints for:

• GET /api/agents/user/agents, /agents/templates/agents, /agents/featured/agents — agent library
• POST /api/agents — create an agent
• GET/PUT /api/agents/by-slug/{slug} — read/update an agent
• POST /api/agents/by-slug/{slug}/chat — chat or generate with an agent
• POST /api/agents/by-slug/{slug}/like — like an agent
• POST /api/agents/suggest — get suggested agents
• GET /api/agents/skills — list available agent skills

These all proxy through to a configurable base URL, so you can point the backend at whichever agent/model provider you want to power generation and chat.

Tech stack

• Frontend: Next.js 16, React 19, Tailwind CSS 4, React Flow
• Backend: FastAPI, SQLAlchemy, asyncpg, httpx
• Monorepo: npm workspaces (client, server, packages/agents)

Status

Early work in progress. Contributions welcome — see the routers and client agents/ pages for the current surface area.

Quick start

Backend

cd server
pip install -r requirements.txt
cp .env .env.local   # set MUAPI_BASE_URL and your API key
uvicorn app.main:app --reload

Frontend

cd client
npm install
npm run dev

Then open http://localhost:3000. The API runs on port 8000 by default; the client is configured to talk to it via CORS.

Related Projects

• MuAPI — Unified API for image, video, and audio generation across hundreds of AI models. Explore AI agents and the model playground.
• Open-Pomelli — Open-source Pomelli alternative — self-hosted AI marketing assistant
• open-character-ai — Open-source Character.AI alternative with custom AI personas

License

MIT