{"id":"autogpt","name":"autogpt-agents","summary":"連続エージェントの構築と展開のための自律型AIエージェントプラットフォーム。ビジュアルワークフローエージェントの作成、持続的な自律エージェントの展開、複雑な多段階AI自動化システムの構築などに利用されます。","body":"# AutoGPT - Autonomous AI Agent Platform\n\nComprehensive platform for building, deploying, and managing continuous AI agents through a visual interface or development toolkit.\n\n## When to use AutoGPT\n\n**Use AutoGPT when:**\n- Building autonomous agents that run continuously\n- Creating visual workflow-based AI agents\n- Deploying agents with external triggers (webhooks, schedules)\n- Building complex multi-step automation pipelines\n- Need a no-code/low-code agent builder\n\n**Key features:**\n- **Visual Agent Builder**: Drag-and-drop node-based workflow editor\n- **Continuous Execution**: Agents run persistently with triggers\n- **Marketplace**: Pre-built agents and blocks to share/reuse\n- **Block System**: Modular components for LLM, tools, integrations\n- **Forge Toolkit**: Developer tools for custom agent creation\n- **Benchmark System**: Standardized agent performance testing\n\n**Use alternatives instead:**\n- **LangChain/LlamaIndex**: If you need more control over agent logic\n- **CrewAI**: For role-based multi-agent collaboration\n- **OpenAI Assistants**: For simple hosted agent deployments\n- **Semantic Kernel**: For Microsoft ecosystem integration\n\n## Quick start\n\n### Installation (Docker)\n\n```bash\n# Clone repository\ngit clone https://github.com/Significant-Gravitas/AutoGPT.git\ncd AutoGPT/autogpt_platform\n\n# Copy environment file\ncp .env.example .env\n\n# Start backend services\ndocker compose up -d --build\n\n# Start frontend (in separate terminal)\ncd frontend\ncp .env.example .env\nnpm install\nnpm run dev\n```\n\n### Access the platform\n\n- **Frontend UI**: http://localhost:3000\n- **Backend API**: http://localhost:8006/api\n- **WebSocket**: ws://localhost:8001/ws\n\n## Architecture overview\n\nAutoGPT has two main systems:\n\n### AutoGPT Platform (Production)\n- Visual agent builder with React frontend\n- FastAPI backend with execution engine\n- PostgreSQL + Redis + RabbitMQ infrastructure\n\n### AutoGPT Classic (Development)\n- **Forge**: Agent development toolkit\n- **Benchmark**: Performance testing framework\n- **CLI**: Command-line interface for development\n\n## Core concepts\n\n### Graphs and nodes\n\nAgents are represented as **graphs** containing **nodes** connected by **links**:\n\n```\nGraph (Agent)\n  ├── Node (Input)\n  │   └── Block (AgentInputBlock)\n  ├── Node (Process)\n  │   └── Block (LLMBlock)\n  ├── Node (Decision)\n  │   └── Block (SmartDecisionMaker)\n  └── Node (Output)\n      └── Block (AgentOutputBlock)\n```\n\n### Blocks\n\nBlocks are reusable functional components:\n\n| Block Type | Purpose |\n|------------|---------|\n| `INPUT` | Agent entry points |\n| `OUTPUT` | Agent outputs |\n| `AI` | LLM calls, text generation |\n| `WEBHOOK` | External triggers |\n| `STANDARD` | General operations |\n| `AGENT` | Nested agent execution |\n\n### Execution flow\n\n```\nUser/Trigger → Graph Execution → Node Execution → Block.execute()\n     ↓              ↓                 ↓\n  Inputs      Queue System      Output Yields\n```\n\n## Building agents\n\n### Using the visual builder\n\n1. **Open Agent Builder** at http://localhost:3000\n2. **Add blocks** from the BlocksControl panel\n3. **Connect nodes** by dragging between handles\n4. **Configure inputs** in each node\n5. **Run agent** using PrimaryActionBar\n\n### Available blocks\n\n**AI Blocks:**\n- `AITextGeneratorBlock` - Generate text with LLMs\n- `AIConversationBlock` - Multi-turn conversations\n- `SmartDecisionMakerBlock` - Conditional logic\n\n**Integration Blocks:**\n- GitHub, Google, Discord, Notion connectors\n- Webhook triggers and handlers\n- HTTP request blocks\n\n**Control Blocks:**\n- Input/Output blocks\n- Branching and decision nodes\n- Loop and iteration blocks\n\n## Agent execution\n\n### Trigger types\n\n**Manual execution:**\n```http\nPOST /api/v1/graphs/{graph_id}/execute\nContent-Type: application/json\n\n{\n  \"inputs\": {\n    \"input_name\": \"value\"\n  }\n}\n```\n\n**Webhook trigger:**\n```http\nPOST /api/v1/webhooks/{webhook_id}\nContent-Type: application/json\n\n{\n  \"data\": \"webhook payload\"\n}\n```\n\n**Scheduled execution:**\n```json\n{\n  \"schedule\": \"0 */2 * * *\",\n  \"graph_id\": \"graph-uuid\",\n  \"inputs\": {}\n}\n```\n\n### Monitoring execution\n\n**WebSocket updates:**\n```javascript\nconst ws = new WebSocket('ws://localhost:8001/ws');\n\nws.onmessage = (event) => {\n  const update = JSON.parse(event.data);\n  console.log(`Node ${update.node_id}: ${update.status}`);\n};\n```\n\n**REST API polling:**\n```http\nGET /api/v1/executions/{execution_id}\n```\n\n## Using Forge (Development)\n\n### Create custom agent\n\n```bash\n# Setup forge environment\ncd classic\n./run setup\n\n# Create new agent from template\n./run forge create my-agent\n\n# Start agent server\n./run forge start my-agent\n```\n\n### Agent structure\n\n```\nmy-agent/\n├── agent.py          # Main agent logic\n├── abilities/        # Custom abilities\n│   ├── __init__.py\n│   └── custom.py\n├── prompts/          # Prompt templates\n└── config.yaml       # Agent configuration\n```\n\n### Implement custom ability\n\n```python\nfrom forge import Ability, ability\n\n@ability(\n    name=\"custom_search\",\n    description=\"Search for information\",\n    parameters={\n        \"query\": {\"type\": \"string\", \"description\": \"Search query\"}\n    }\n)\ndef custom_search(query: str) -> str:\n    \"\"\"Custom search ability.\"\"\"\n    # Implement search logic\n    result = perform_search(query)\n    return result\n```\n\n## Benchmarking agents\n\n### Run benchmarks\n\n```bash\n# Run all benchmarks\n./run benchmark\n\n# Run specific category\n./run benchmark --category coding\n\n# Run with specific agent\n./run benchmark --agent my-agent\n```\n\n### Benchmark categories\n\n- **Coding**: Code generation and debugging\n- **Retrieval**: Information finding\n- **Web**: Web browsing and interaction\n- **Writing**: Text generation tasks\n\n### VCR cassettes\n\nBenchmarks use recorded HTTP responses for reproducibility:\n\n```bash\n# Record new cassettes\n./run benchmark --record\n\n# Run with existing cassettes\n./run benchmark --playback\n```\n\n## Integrations\n\n### Adding credentials\n\n1. Navigate to Profile > Integrations\n2. Select provider (OpenAI, GitHub, Google, etc.)\n3. Enter API keys or authorize OAuth\n4. Credentials are encrypted and stored securely\n\n### Using credentials in blocks\n\nBlocks automatically access user credentials:\n\n```python\nclass MyLLMBlock(Block):\n    def execute(self, inputs):\n        # Credentials are injected by the system\n        credentials = self.get_credentials(\"openai\")\n        client = OpenAI(api_key=credentials.api_key)\n        # ...\n```\n\n### Supported providers\n\n| Provider | Auth Type | Use Cases |\n|----------|-----------|-----------|\n| OpenAI | API Key | LLM, embeddings |\n| Anthropic | API Key | Claude models |\n| GitHub | OAuth | Code, repos |\n| Google | OAuth | Drive, Gmail, Calendar |\n| Discord | Bot Token | Messaging |\n| Notion | OAuth | Documents |\n\n## Deployment\n\n### Docker production setup\n\n```yaml\n# docker-compose.prod.yml\nservices:\n  rest_server:\n    image: autogpt/platform-backend\n    environment:\n      - DATABASE_URL=postgresql://...\n      - REDIS_URL=redis://redis:6379\n    ports:\n      - \"8006:8006\"\n\n  executor:\n    image: autogpt/platform-backend\n    command: poetry run executor\n\n  frontend:\n    image: autogpt/platform-frontend\n    ports:\n      - \"3000:3000\"\n```\n\n### Environment variables\n\n| Variable | Purpose |\n|----------|---------|\n| `DATABASE_URL` | PostgreSQL connection |\n| `REDIS_URL` | Redis connection |\n| `RABBITMQ_URL` | RabbitMQ connection |\n| `ENCRYPTION_KEY` | Credential encryption |\n| `SUPABASE_URL` | Authentication |\n\n### Generate encryption key\n\n```bash\ncd autogpt_platform/backend\npoetry run cli gen-encrypt-key\n```\n\n## Best practices\n\n1. **Start simple**: Begin with 3-5 node agents\n2. **Test incrementally**: Run and test after each change\n3. **Use webhooks**: External triggers for event-driven agents\n4. **Monitor costs**: Track LLM API usage via credits system\n5. **Version agents**: Save working versions before changes\n6. **Benchmark**: Use agbenchmark to validate agent quality\n\n## Common issues\n\n**Services not starting:**\n```bash\n# Check container status\ndocker compose ps\n\n# View logs\ndocker compose logs rest_server\n\n# Restart services\ndocker compose restart\n```\n\n**Database connection issues:**\n```bash\n# Run migrations\ncd backend\npoetry run prisma migrate deploy\n```\n\n**Agent execution stuck:**\n```bash\n# Check RabbitMQ queue\n# Visit http://localhost:15672 (guest/guest)\n\n# Clear stuck executions\ndocker compose restart executor\n```\n\n## References\n\n- **[Advanced Usage](references/advanced-usage.md)** - Custom blocks, deployment, scaling\n- **[Troubleshooting](references/troubleshooting.md)** - Common issues, debugging\n\n## Resources\n\n- **Documentation**: https://docs.agpt.co\n- **Repository**: https://github.com/Significant-Gravitas/AutoGPT\n- **Discord**: https://discord.gg/autogpt\n- **License**: MIT (Classic) / Polyform Shield (Platform)","author":"@Orchestra-Research","ownerProfile":null,"authorContacts":null,"sourceUrl":"https://github.com/Orchestra-Research/AI-Research-SKILLs/tree/main/14-agents/autogpt","license":"MIT","category":"writing","lang":"en","tokens":2137,"stars":0,"calls30d":1,"claimed":false,"visibility":"public","origin":"crawler","version":"0.1.0","createdAt":"2026-08-22","updatedAt":"2026-08-22","files":[{"path":"references/advanced-usage.md","size":13209,"sha256":"e0b9f472e7487061b5123f1a507ebd1fa19bb0e91f95de5931c7a5b6279b104b"},{"path":"references/troubleshooting.md","size":7689,"sha256":"af763d0e0c0a0120a8b2ab2534af34697b8d8183c907c271e6783bf5c6f61500"}],"requires":{"mcp":[],"tools":[]},"safety":{"flags":[],"scannedAt":"2026-08-22","hasScripts":false,"networkEndpoints":["discord.gg","docs.agpt.co"]}}