OpenAgents 开源自托管 AI 智能体平台发布

The open-source, self-hosted AI agent platform.
Build, run, and own your own AI agents — no cloud required.

🌐 Website • 🚀 Install • 💡 Features • 🏢 Plans

What is OpenAgents?

OpenAgents is a self-hosted AI assistant platform for complex, long-horizon tasks through multi-agent collaboration. It supports content creation, coding, research, image generation, and more — all running on your own machine or server.

Installation

Recommended path: use the guided setup for local development, and Docker Compose for production/self-hosting.
First launch can take 5–10 minutes depending on image builds and model setup.

Quick install checklist

Before starting, make sure you have:

• Node.js 20+
• pnpm 9 (via Corepack — no separate download needed, see OS steps below)
• Docker Desktop / Docker Engine
• Git

For a fully functional install, OpenAgents also needs:

• PostgreSQL and Redis running locally or via Docker
• At least one model provider configured after first login:
  • Ollama for local/self-hosted usage, or
  • OpenAI / Anthropic / Gemini / other supported API key

Recommended local setup

From the repo root, run:

pnpm setup
pnpm dev:watch

pnpm dev:watch is the self-healing server: it restarts the API/web if they crash and revives Ollama when its port goes dark (plain pnpm dev still works too).

What pnpm setup does:

• installs dependencies
• creates missing .env files from examples
• starts local Postgres + Redis with Docker when available
• runs Prisma generate + migrations

After startup:

• Web: http://localhost:3000/login
• API health: http://localhost:3001/api/v1/health

If ports 3000/3001 are already taken (e.g. another app), start with alternates instead of pnpm dev:

WEB_PORT=3002 API_PORT=3101 NEXT_PUBLIC_API_URL=http://localhost:3101 pnpm dev

then open http://localhost:3002/login.

Then log in and go to Settings → Config to add your provider key or connect Ollama.

Install as a desktop app

Once OpenAgents is running:

1. Open http://localhost:3000 in Chrome or Edge and sign in
2. Browser menu → Apps → Install this site as an app
3. Pin the OpenAgents icon — it launches in its own window with its own icon

The one-shot installers (scripts/install.ps1, scripts/install.sh) do the full job hands-free: prerequisites, optional Ollama (-InstallOllama / --with-ollama), a self-healing launcher that revives the app and Ollama if either stops (auto-starts at sign-in), plus a desktop shortcut and OPENAGENTS-START-HERE.txt.

🪟 Windows

Step 1 — Install Docker Desktop

Download and install Docker Desktop. During install, check "Use WSL 2 instead of Hyper-V".

👉 Download Docker Desktop for Windows

Open Docker Desktop and wait until you see "Engine running" in the bottom-left corner before continuing.

Step 2 — Install Git

👉 Download Git for Windows — install with all default settings.

Step 3 — Install Node.js 20 + pnpm

👉 Download Node.js 20 LTS — install, then close and reopen PowerShell so PATH refreshes, and run:

corepack enable
corepack prepare pnpm@9.0.0 --activate
pnpm --version

Step 4 — Open PowerShell and run these commands

Press Win + X → click Terminal or PowerShell, then paste:

git clone https://github.com/edisonmliranzo/openagents.git
cd openagents
pnpm setup

Step 4 — Start OpenAgents

pnpm dev:watch

Step 5 — Open your browser

http://localhost:3000/login

Step 6 — Make it fully functional

• Go to Settings → Config
• Add an API key or run Ollama locally and set its URL
• Confirm API health at http://localhost:3001/api/v1/health

If you want the production Docker stack instead of local dev, use:

cd infra/docker
Copy-Item .env.prod.example .env.prod
docker compose -f docker-compose.prod.yml up --build -d

(Git Bash: cp .env.prod.example .env.prod instead of Copy-Item.)

🍎 macOS

Step 1 — Install Docker Desktop

👉 Download Docker Desktop for Mac

Drag it to your Applications folder, open it, and wait for "Engine running".

Step 2 — Install Node.js 20 + pnpm

👉 Download Node.js 20 LTS — install, then reopen Terminal and run:

corepack enable
corepack prepare pnpm@9.0.0 --activate
pnpm --version

Step 3 — Open Terminal and run these commands

Press Cmd + Space, type Terminal, press Enter, then paste:

git clone https://github.com/edisonmliranzo/openagents.git
cd openagents
pnpm setup

Step 3 — Start OpenAgents

pnpm dev:watch

Step 4 — Open your browser

http://localhost:3000/login

Then go to Settings → Config to add an API key or connect Ollama.

For the production Docker stack instead:

cd infra/docker
cp .env.prod.example .env.prod
docker compose -f docker-compose.prod.yml up --build -d

🟠 Ubuntu

Step 1 — Install Docker (one-liner)

curl -fsSL https://get.docker.com | sh && sudo systemctl enable --now docker

Verify: docker --version

Step 2 — Install Node.js 20 + pnpm

curl -fsSL https://deb.nodesource.com/setup_20.x | sudo -E bash -
sudo apt-get install -y nodejs
corepack enable
corepack prepare pnpm@9.0.0 --activate

Step 3 — Clone the repository

git clone https://github.com/edisonmliranzo/openagents.git
cd openagents

Step 4 — Run guided setup

pnpm setup
pnpm dev:watch

Go to Settings → Config to add your API key or connect Ollama.

Open your browser:

http://YOUR_SERVER_IP:3000/login

🟢 Linux / VPS / Production Docker

Step 1 — Install Docker

curl -fsSL https://get.docker.com | sh && sudo systemctl enable --now docker

Step 2 — Clone and configure

git clone https://github.com/edisonmliranzo/openagents.git
cd openagents
pnpm setup
# pnpm setup creates infra/docker/.env.prod with generated secrets —
# or copy manually: cp infra/docker/.env.prod.example infra/docker/.env.prod

Step 3 — Edit infra/docker/.env.prod before launch

pnpm setup already generated random JWT_SECRET, JWT_REFRESH_SECRET, ENCRYPTION_KEY, and POSTGRES_PASSWORD — verify they are non-placeholder. At minimum, also set:

• one provider setting such as OPENAI_API_KEY or local OLLAMA_BASE_URL
• FRONTEND_URL / NEXT_PUBLIC_API_URL if not using localhost

Step 4 — Launch (from the repo root)

pnpm prod:deploy

Step 5 — Verify the installation

pnpm prod:ps
curl http://localhost:3001/api/v1/health
curl -I http://localhost:3000/login

If Ollama is used from the host machine, make sure it is actually running and reachable from Docker via host.docker.internal.

Open your browser:

http://YOUR_SERVER_IP:3000/login

Post-install: fully functional checklist

Your installation is not fully usable until these are true:

• Web loads at /login
• API health endpoint returns success
• Database migrations completed
• Redis is reachable
• You can log in / register
• At least one LLM provider is configured
• Optional media features have keys set if needed:
  • OPENAI_API_KEY for DALL-E / TTS
  • STABILITY_API_KEY for Stability image generation
  • ELEVENLABS_API_KEY for ElevenLabs audio
• Mobile (Expo): copy apps/mobile/.env.example to apps/mobile/.env and set EXPO_PUBLIC_API_URL to your machine's LAN URL (not localhost — phones can't reach your PC's localhost)

Troubleshooting

• If pnpm setup fails, confirm Node.js 20+ and pnpm 9 (corepack enable then corepack prepare pnpm@9.0.0 --activate).
• If pnpm is not recognized after install, close and reopen the terminal so PATH refreshes.
• If chat opens but the agent never answers, configure a provider in Settings → Config.
• If Ollama is selected, verify the Ollama server is running and the base URL is correct.
• If Docker production healthchecks fail, inspect:

cd infra/docker
docker compose -f docker-compose.prod.yml logs --tail=200 api web worker

Features

🤖 AI Capabilities

🔧 Built-in Tools (50+)

🔌 LLM Providers

Connect any of these in Settings → Config — open a provider card, paste the key, Test, then set it as Active LLM:

OpenAI · Anthropic · Google Gemini · Ollama (local) · MiniMax · Perplexity · NVIDIA NIM · AtlasCloud · Groq · Mistral · DeepSeek · xAI Grok · OpenRouter · Together AI · Meta Muse Spark · Custom endpoint (any OpenAI-compatible /v1 server — LM Studio, vLLM, llama.cpp)

The model picker loads each provider's live model catalog automatically (with a curated fallback), and you can type any model id — including new releases the day they ship.

Architecture

┌─────────────────────────────────────────┐
│            OpenAgents Platform           │
├────────────┬────────────┬───────────────┤
│  Web UI    │    API     │    Worker     │
│  Next.js   │   NestJS   │  Background   │
├────────────┴────────────┴───────────────┤
│              Core Services              │
│  Agent Runtime · Tool Engine · Memory  │
├─────────────────────────────────────────┤
│           Data & Infrastructure         │
│    PostgreSQL · Redis · Docker          │
└─────────────────────────────────────────┘

Plans

All plans run on your own server. You own your data.

Contact: edison0220@gmail.com

Tech Stack

License

OpenAgents is open-source under the MIT License.