LobeChat - Open-source AI agent workspace with multi-provider LLM support, plugin system, knowledge base RAG, 505+ agents, and self-hosting options via Docker/Vercel
LobeChat is an open-source, modern AI agent workspace that provides a unified platform for conversing with multiple AI systems. It supports 20+ LLM providers (OpenAI, Claude, Gemini, Ollama), features a plugin marketplace with 40+ plugins, an agent marketplace with 505+ pre-built agents, knowledge base with RAG, and flexible deployment options.
Key Value Proposition: Deploy a private, customizable AI chat interface with multi-model support, extensible plugins, file upload/RAG capabilities, and beautiful modern UI - all self-hosted or cloud-deployed.
┌─────────────────────────────────────────────────────────────────┐
│ LobeChat Workspace │
│ (Next.js + TypeScript) │
└─────────────────────────────────────────────────────────────────┘
│
┌─────────────────────┼─────────────────────┐
│ │ │
▼ ▼ ▼
┌───────────────┐ ┌───────────────┐ ┌───────────────┐
│ AI Providers │ │ Agents │ │ Plugins │
├───────────────┤ ├───────────────┤ ├───────────────┤
│ • OpenAI │ │ • 505+ pre- │ │ • 40+ plugins │
│ • Claude │ │ built │ │ • Web search │
│ • Gemini │ │ • Custom │ │ • Code exec │
│ • Ollama │ │ • Marketplace │ │ • Weather │
│ • Azure │ │ • System │ │ • Custom SDK │
│ • Groq │ │ prompts │ │ │
│ • 20+ more │ │ │ │ │
└───────────────┘ └───────────────┘ └───────────────┘
│ │ │
└─────────────────────┼─────────────────────┘
│
┌─────────────────────┼─────────────────────┐
▼ ▼ ▼
┌───────────────┐ ┌───────────────┐ ┌───────────────┐
│ Knowledge Base│ │ Features │ │ Deployment │
├───────────────┤ ├───────────────┤ ├───────────────┤
│ • File upload │ │ • TTS/STT │ │ • Docker │
│ • RAG │ │ • Vision │ │ • Vercel │
│ • pgvector │ │ • Artifacts │ │ • Alibaba │
│ • Embeddings │ │ • DALL-E │ │ • Zeabur │
│ │ │ • MCP │ │ • Self-host │
└───────────────┘ └───────────────┘ └───────────────┘
| Metric | Value |
|---|---|
| GitHub Stars | 70.1k+ |
| Forks | 14.4k+ |
| Contributors | 313+ |
| Total Commits | 8,248+ |
| Plugins | 40+ |
| Agents | 505+ |
| Languages | TypeScript (98.5%) |
| Provider | Models | Notes |
|---|---|---|
| OpenAI | GPT-4o, GPT-4, GPT-3.5 | Primary provider |
| Anthropic | Claude 3.5, Claude 3 | Full model support |
| Gemini Pro, Gemini Flash | Vision support | |
| Azure OpenAI | All OpenAI models | Enterprise option |
| Ollama | Llama, Mistral, etc. | Local models |
| Groq | Llama, Mixtral | Fast inference |
| Moonshot | Kimi models | Chinese provider |
| 01 AI | Yi models | Chinese provider |
| Together AI | Open source models | Many models |
| ChatGLM | GLM-4 | Chinese provider |
| Perplexity | pplx-* models | Search-enhanced |
| Mistral | Mistral models | European provider |
| DeepSeek | DeepSeek models | Coding focused |
# Deploy to Vercel (simplest option)
# Click "Deploy" button on GitHub repo or:
npx vercel --prod
# Quick start with local storage
docker run -d \
--name lobe-chat \
-p 3210:3210 \
-e OPENAI_API_KEY=sk-your-key \
lobehub/lobe-chat
# Full deployment with PostgreSQL
docker run -d \
--name lobe-chat-db \
-p 3210:3210 \
-e DATABASE_URL=postgres://user:pass@host:5432/lobechat \
-e KEY_VAULTS_SECRET=$(openssl rand -base64 32) \
-e OPENAI_API_KEY=sk-your-key \
lobehub/lobe-chat-database
# docker-compose.yml
version: '3.8'
services:
lobe-chat:
image: lobehub/lobe-chat-database
ports:
- "3210:3210"
environment:
- DATABASE_URL=postgres://postgres:password@db:5432/lobechat
- KEY_VAULTS_SECRET=${KEY_VAULTS_SECRET}
- OPENAI_API_KEY=${OPENAI_API_KEY}
- NEXT_AUTH_SECRET=${NEXT_AUTH_SECRET}
depends_on:
- db
db:
image: pgvector/pgvector:pg16
environment:
- POSTGRES_USER=postgres
- POSTGRES_PASSWORD=password
- POSTGRES_DB=lobechat
volumes:
- postgres_data:/var/lib/postgresql/data
volumes:
postgres_data:
# Generate secrets
export KEY_VAULTS_SECRET=$(openssl rand -base64 32)
export NEXT_AUTH_SECRET=$(openssl rand -base64 32)
export OPENAI_API_KEY=sk-your-key
# Start services
docker-compose up -d
| Variable | Description | Required |
|---|---|---|
OPENAI_API_KEY |
OpenAI API key | Yes (or other provider) |
ACCESS_CODE |
Password protection | No |
PROXY_URL |
Proxy for API requests | No |
| Variable | Description | Default |
|---|---|---|
NEXT_PUBLIC_SERVICE_MODE |
client or server |
client |
DATABASE_URL |
PostgreSQL connection string | - |
DATABASE_DRIVER |
node or neon |
node |
KEY_VAULTS_SECRET |
Encryption key for user data | Required for server mode |
# Set provider keys as needed
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GOOGLE_API_KEY=...
AZURE_API_KEY=...
GROQ_API_KEY=gsk_...
OLLAMA_PROXY_URL=http://localhost:11434
| Variable | Description | Example |
|---|---|---|
FEATURE_FLAGS |
Enable/disable features | +webSearch,-imageGen |
DEFAULT_AGENT_CONFIG |
Default agent settings | JSON config |
SYSTEM_AGENT |
System function models | JSON config |
PLUGINS_INDEX_URL |
Custom plugin marketplace | URL |
AGENTS_INDEX_URL |
Custom agent marketplace | URL |
# NextAuth (Docker)
NEXT_AUTH_SECRET=$(openssl rand -base64 32)
NEXTAUTH_URL=https://your-domain.com
# Clerk (Vercel)
CLERK_SECRET_KEY=sk_...
NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY=pk_...
{
"meta": {
"name": "Code Assistant",
"description": "Expert programming assistant",
"avatar": "🤖",
"tags": ["coding", "development"]
},
"config": {
"systemRole": "You are an expert programmer...",
"model": "gpt-4o",
"params": {
"temperature": 0.3,
"max_tokens": 4096
},
"plugins": ["web-search", "code-interpreter"],
"tts": {
"enabled": true,
"voice": "alloy"
}
}
}
# Role Definition
You are [role name], an expert in [domain].
# Core Competencies
- [Skill 1]
- [Skill 2]
- [Skill 3]
# Guidelines
1. [Guideline 1]
2. [Guideline 2]
# Response Format
- Use markdown formatting
- Include code examples when relevant
- Cite sources when possible
# Constraints
- Do not [constraint 1]
- Always [requirement 1]
agents/ directoryPlugin
├── manifest.json # Plugin definition
├── api/
│ └── endpoint.ts # Server-side handlers
└── ui/
└── index.tsx # Optional frontend UI
# Clone official template
git clone https://github.com/lobehub/chat-plugin-template.git
cd chat-plugin-template
npm install
npm run dev
# Runs on http://localhost:3400
{
"identifier": "weather-plugin",
"version": "1.0.0",
"api": [
{
"url": "http://localhost:3400/api/weather",
"name": "getWeather",
"description": "Get current weather for a location",
"parameters": {
"type": "object",
"properties": {
"city": {
"type": "string",
"description": "City name"
}
},
"required": ["city"]
}
}
],
"ui": {
"url": "http://localhost:3400",
"height": 400,
"width": 600
},
"gateway": "http://localhost:3400/api/gateway"
}
// api/weather.ts
import { createErrorResponse } from '@lobehub/chat-plugin-sdk';
export const runtime = 'edge';
export async function POST(req: Request) {
try {
const { city } = await req.json();
// Fetch weather data
const response = await fetch(
`https://api.weather.com/v1/current?city=${city}`
);
const data = await response.json();
return Response.json({
temperature: data.temp,
condition: data.condition,
city: city
});
} catch (error) {
return createErrorResponse(500, 'Failed to fetch weather');
}
}
// ui/index.tsx
import { fetchPluginMessage } from '@lobehub/chat-plugin-sdk/client';
export default function PluginUI() {
const [message, setMessage] = useState(null);
useEffect(() => {
fetchPluginMessage().then(setMessage);
}, []);
return (
<div>
{message && <pre>{JSON.stringify(message, null, 2)}</pre>}
</div>
);
}
# Install SDK
npm install @lobehub/chat-plugin-sdk @lobehub/chat-plugins-gateway
// Using gateway for local development
import { createGateway } from '@lobehub/chat-plugins-gateway';
const gateway = createGateway({
plugins: ['./manifest.json']
});
-- Enable pgvector
CREATE EXTENSION IF NOT EXISTS vector;
-- Vectors stored automatically by LobeChat
# Embedding provider (optional, uses OpenAI by default)
EMBEDDING_MODEL=text-embedding-3-small
# Vector search settings
VECTOR_SEARCH_LIMIT=10
VECTOR_SEARCH_THRESHOLD=0.7
LobeChat supports MCP for connecting to external tools and data sources.
# Environment variable
MCP_ENABLED=true
{
"mcpServers": {
"filesystem": {
"command": "npx",
"args": ["-y", "@anthropic-ai/mcp-server-filesystem", "/path/to/files"]
},
"github": {
"command": "npx",
"args": ["-y", "@anthropic-ai/mcp-server-github"],
"env": {
"GITHUB_TOKEN": "your-token"
}
}
}
}
# Enable TTS
TTS_ENABLED=true
TTS_PROVIDER=openai # or edge, elevenlabs
# Enable STT
STT_ENABLED=true
STT_PROVIDER=openai # or browser
# DALL-E
DALL_E_ENABLED=true
# Image settings
AI_IMAGE_DEFAULT_IMAGE_NUM=4
Artifacts enable real-time creation and visualization of:
Visualize model reasoning steps for transparency in complex tasks.
Create non-linear dialogue trees, exploring multiple conversation paths.
# Basic
docker pull lobehub/lobe-chat
docker run -p 3210:3210 lobehub/lobe-chat
# With database
docker pull lobehub/lobe-chat-database
docker run -p 3210:3210 \
-e DATABASE_URL=... \
lobehub/lobe-chat-database
apiVersion: apps/v1
kind: Deployment
metadata:
name: lobe-chat
spec:
replicas: 3
selector:
matchLabels:
app: lobe-chat
template:
metadata:
labels:
app: lobe-chat
spec:
containers:
- name: lobe-chat
image: lobehub/lobe-chat-database
ports:
- containerPort: 3210
env:
- name: DATABASE_URL
valueFrom:
secretKeyRef:
name: lobe-secrets
key: database-url
# Required
NEXT_AUTH_SECRET=$(openssl rand -base64 32)
NEXTAUTH_URL=https://your-domain.com
# OAuth Providers
GITHUB_ID=...
GITHUB_SECRET=...
GOOGLE_ID=...
GOOGLE_SECRET=...
CLERK_SECRET_KEY=sk_...
NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY=pk_...
CLERK_WEBHOOK_SECRET=whsec_...
# Password protection
ACCESS_CODE=your-secret-code
Error: Failed to connect to OpenAI
Solutions:
PROXY_URL if behind firewallError: Connection refused to PostgreSQL
Solutions:
DATABASE_URL formatError: Plugin manifest not found
Solutions:
Error: Vector search failed
Solutions:
Error: NextAuth session invalid
Solutions:
NEXT_AUTH_SECRETNEXTAUTH_URL matches domainACCESS_CODE for public deploymentsopenssl rand -base64 32NEXT_PUBLIC_ASSET_PREFIX