💡 摘要
AgentStack是一个多智能体框架,提供广泛的金融智能和AI编排工具。
🎯 适合人群
🤖 AI 吐槽: “这就像是AI代理的瑞士军刀,但我希望它不会带着钝刀。”
该框架可能暴露与网络访问和依赖供应链漏洞相关的风险。实施严格的依赖管理和定期安全审计可以减轻这些风险。
🚀 AgentStack

AgentStack is a production-grade multi-agent framework built on Mastra, delivering 60+ enterprise tools, 31+ specialized agents, 15 workflows, 13 agent networks, 65 UI components (30+ AI Elements + 35+ base), and A2A/MCP orchestration for scalable AI systems. Focuses on financial intelligence, RAG pipelines, observability, secure governance, and AI chat interfaces.
🎯 Why AgentStack?
| Feature | AgentStack | LangChain | CrewAI | AutoGen | | ---------------------------- | ---------------------------------------------------- | ------------- | ------------- | ---------- | | Production Observability | ✅ Full Langfuse tracing + custom scorers | ⚠️ Partial | ❌ Basic | ❌ Limited | | Financial Tools | ✅ Polygon/Finnhub/AlphaVantage (30+ endpoints) | ❌ None | ❌ None | ❌ None | | RAG Pipeline | ✅ PgVector HNSW + rerank + graphRAG | ⚠️ External | ❌ Basic | ❌ None | | Multi-Agent | ✅ A2A MCP + parallel orchestration (30+ agents) | ⚠️ Sequential | ✅ Sequential | ✅ Custom | | Governance | ✅ JWT/RBAC + path traversal + HTML sanitization | ❌ Custom | ❌ None | ❌ None | | TypeScript | ✅ Zod schemas everywhere (94+ tools) | ⚠️ JS/TS mix | ⚠️ JS focus | ❌ Python | | UI Components | ✅ 65 components (AI Elements + shadcn/ui) | ❌ None | ❌ None | ❌ None | | Tests | ✅ Vitest + comprehensive test suite | ⚠️ Partial | ❌ Sparse | ⚠️ Partial |
Built for production: Secure, observable, testable agents with zero-config PgVector RAG + enterprise financial APIs.
✨ Core Capabilities
- 💰 Financial Intelligence: 30+ tools (Polygon quotes/aggs/fundamentals, Finnhub analysis, AlphaVantage indicators)
- 🔍 Semantic RAG: PgVector (3072D embeddings) + MDocument chunking + rerank + graph traversal
- 🤖 31+ Agents: Research → Learn → Report → Edit → Analyze (stock/crypto/copywriter/evaluator/data pipeline/business-legal/charting/image/coding/dane/social media/SEO/translation/customer support/project management)
- 📋 15 Workflows: Weather, content, financial reports, document processing, research synthesis, learning extraction, governed RAG (index + answer), spec generation, repo ingestion, stock analysis, marketing campaign
- 🌐 13 Agent Networks: Primary routing, data pipeline, report generation, research pipeline, content creation, financial intelligence, learning, marketing automation, DevOps, business intelligence, security
- 🔌 A2A/MCP: MCP server coordinates parallel agents (research+stock→report), A2A coordinator for cross-agent communication
- 🎨 65 UI Components: AI Elements (30 chat/reasoning/canvas components) + shadcn/ui (35 base primitives)
- 📊 Full Observability: Langfuse traces + 10+ custom scorers (diversity/quality/completeness) + TanStack Query for state management
- 🛡️ Enterprise Security: JWT auth, RBAC, path validation, HTML sanitization, secrets masking
- ⚡ Extensible: Model registry (Gemini/OpenAI/Anthropic/OpenRouter), Zod schemas everywhere, MastraClient SDK integration
🏗️ Architecture
%%{init: {'theme': 'dark', 'themeVariables': { 'primaryColor': '#58a6ff', 'primaryTextColor': '#c9d1d9', 'primaryBorderColor': '#30363d', 'lineColor': '#58a6ff', 'sectionBkgColor': '#161b22', 'altSectionBkgColor': '#0d1117', 'sectionTextColor': '#c9d1d9', 'gridColor': '#30363d', 'tertiaryColor': '#161b22' }}}%% graph TB subgraph "� Frontend (Next.js 16)" UI[AI Elements + shadcn/ui<br/>• 30+ AI Components<br/>• 35+ Base Primitives] App[App Router<br/>• React 19<br/>• Tailwind CSS 4] end subgraph "🌐 MCP/A2A Client" Client[Cursor/Claude/External Agents] --> Coord[A2A Coordinator MCP] end subgraph "🎯 AgentStack Runtime" Coord --> Agents[30+ Agents<br/>• Research/Financial/Coding<br/>• Content/Data/Business] Agents --> Tools[60+ Tools<br/>• Polygon/Finnhub/SerpAPI<br/>• RAG/Code/Data Processing] Agents --> Workflows[14+ Workflows<br/>• Weather/Content/Financial<br/>• Document/RAG/Analysis] Agents --> Networks[4+ Networks<br/>• Coding/Data/Report/Research] end subgraph "🗄️ PgVector Storage" Tools --> Embeddings[3072D Gemini<br/>HNSW/Flat Indexes] Tools --> Postgres[Traces/Evals<br/>Memory/Threads] end subgraph "📊 Observability" Agents --> Otel[Otel Traces<br/>• 97% Traced<br/>• 10+ Scorers] UI --> Otel Postgres --> Otel Agents --> Otel Coord --> Otel Workflows --> Otel Networks --> Otel end UI --> App App --> Agents style Client stroke:#58a6ff style App stroke:#58a6ff style Coord stroke:#58a6ff style Agents stroke:#58a6ff style Tools stroke:#58a6ff style Embeddings stroke:#58a6ff style Workflows stroke:#58a6ff style Networks stroke:#58a6ff style Postgres stroke:#58a6ff style Otel stroke:#58a6ff style UI stroke:#58a6ff
🔍 Chat UI-Backend Architecture
%%{init: {'theme': 'dark', 'themeVariables': { 'primaryColor': '#58a6ff', 'primaryTextColor': '#c9d1d9', 'primaryBorderColor': '#30363d', 'lineColor': '#58a6ff', 'sectionBkgColor': '#161b22', 'altSectionBkgColor': '#0d1117', 'sectionTextColor': '#c9d1d9', 'gridColor': '#30363d', 'tertiaryColor': '#161b22' }}}%% sequenceDiagram participant UI as ChatUI participant Msg as MessageItem participant TG as TypeGuards participant ADS as AgentDataSection participant WDS as WorkflowDataSection participant NDS as NetworkDataSection parti
优点
- 全面的工具和代理集
- 强烈关注金融智能
- 强大的可观察性和安全性功能
缺点
- 复杂性可能让新用户感到不知所措
- 可能陡峭的学习曲线
- 依赖多个外部库
相关技能
免责声明:本内容来源于 GitHub 开源项目,仅供展示和评分分析使用。
版权归原作者所有 ssdeanx.
