{"id":51215800,"url":"https://github.com/DemonDamon/AgenticX","last_synced_at":"2026-07-16T14:01:13.836Z","repository":{"id":263237273,"uuid":"772408997","full_name":"DemonDamon/AgenticX","owner":"DemonDamon","description":"AgenticX is a unified, production-ready multi-agent platform — Python SDK + CLI (agx) + Studio server + Machi desktop app. Features Meta-Agent orchestration, 15+ LLM providers, MCP Hub, hierarchical memory, avatar \u0026 group chat, skill ecosystem, safety sandbox, and IM gateway (Feishu/WeChat).","archived":false,"fork":false,"pushed_at":"2026-07-16T01:47:41.000Z","size":115733,"stargazers_count":183,"open_issues_count":16,"forks_count":33,"subscribers_count":6,"default_branch":"main","last_synced_at":"2026-07-16T02:21:12.873Z","etag":null,"topics":["agent-framework","agentic-workflows","ai-agent","ai-orchestration","chatbot","desktop-app","electron","fastapi","graphrag","llm","mcp","multi-agent-system","python","rag","tool-use"],"latest_commit_sha":null,"homepage":"https://www.agxbuilder.com/","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/DemonDamon.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":".github/FUNDING.yml","license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":"docs/roadmap.md","authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":"AGENTS.md","dco":null,"cla":null},"funding":{"github":"DemonDamon"}},"created_at":"2024-03-15T06:11:32.000Z","updated_at":"2026-07-16T01:47:46.000Z","dependencies_parsed_at":"2024-11-17T07:32:14.798Z","dependency_job_id":"7c836ae2-9932-4418-950f-adb738ad46e8","html_url":"https://github.com/DemonDamon/AgenticX","commit_stats":null,"previous_names":["demondamon/agentpalace","demondamon/agenticx"],"tags_count":10,"template":false,"template_full_name":null,"purl":"pkg:github/DemonDamon/AgenticX","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/DemonDamon%2FAgenticX","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/DemonDamon%2FAgenticX/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/DemonDamon%2FAgenticX/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/DemonDamon%2FAgenticX/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/DemonDamon","download_url":"https://codeload.github.com/DemonDamon/AgenticX/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/DemonDamon%2FAgenticX/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":35546267,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-07-16T02:00:06.687Z","response_time":83,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["agent-framework","agentic-workflows","ai-agent","ai-orchestration","chatbot","desktop-app","electron","fastapi","graphrag","llm","mcp","multi-agent-system","python","rag","tool-use"],"created_at":"2026-06-28T04:01:00.194Z","updated_at":"2026-07-16T14:01:13.830Z","avatar_url":"https://github.com/DemonDamon.png","language":"Python","funding_links":["https://github.com/sponsors/DemonDamon"],"categories":["⭐ 精选第三方技能"],"sub_categories":["📣 平台运营 / 自媒体 / 办公流量"],"readme":"# AgenticX: Unified Multi-Agent Framework\n\n\u003cdiv align=\"center\"\u003e\n\u003c!-- \u003cimg src=\"assets/agenticx-logo-2025.png\" alt=\"AgenticX Logo\" width=\"240\" style=\"margin-bottom:20px;\" /\u003e --\u003e\n\u003cimg src=\"assets/agenticx-logo-2025.png\" alt=\"AgenticX Logo\" width=\"800\" style=\"margin-bottom:20px;\" /\u003e\n\n\n\u003c!-- [![Python](https://img.shields.io/badge/Python-3.10+-blue.svg)](https://www.python.org/downloads/) --\u003e\n[![License: Apache-2.0](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](https://www.apache.org/licenses/LICENSE-2.0)\n[![PyPI version](https://img.shields.io/pypi/v/agenticx)](https://pypi.org/project/agenticx/)\n[![PyPI - Python Version](https://img.shields.io/pypi/pyversions/agenticx)](https://pypi.org/project/agenticx/)\n[![Ask DeepWiki](https://deepwiki.com/badge.svg)](https://deepwiki.com/DemonDamon/AgenticX)\n\n[Architecture](#system-architecture) • [Features](#core-features) • [Quick Start](#quick-start) • [Examples](#complete-examples) • [Progress](#development-progress)\n\n\u003c/div\u003e\n\n---\n\n**Language / 语言**: [English](README.md) | [中文](README_ZN.md)\n\n---\n\n## Security advisory\n\n**LiteLLM (PyPI):** Malicious releases **`litellm` 1.82.7 and 1.82.8** were removed from PyPI after reports that they could **exfiltrate API keys**. If you ever installed either version, **uninstall** them, **rotate any credentials** that may have been exposed, and **upgrade** to a release the upstream project and PyPI list as safe (for example **1.82.9+**, per current upstream guidance). Check your environment with `pip show litellm`.\n\n---\n\n## Vision\n\n**AgenticX** aims to create a unified, scalable, production-ready multi-agent application development framework, empowering developers to build everything from simple automation assistants to complex collaborative intelligent agent systems.\n\n## System Architecture\n\n\u003cdiv align=\"center\"\u003e\n\u003cimg src=\"assets/AgenticX System Architecture.png\" alt=\"AgenticX System Architecture — 5-tier overview covering UI, Studio Runtime, Core Framework, Platform Services, and Domain Extensions\" width=\"900\" /\u003e\n\u003c/div\u003e\n\nThe framework is organized into 5 tiers: **User Interface** (Desktop / CLI / SDK) → **Studio Runtime** (Session Manager, Meta-Agent, Team Manager, Avatar \u0026 Group Chat) → **Core Framework** (Orchestration, Execution, Agent, Memory, Tools, LLM Providers, Hooks) → **Platform Services** (Observability, Protocols, Security, Storage) → **Domain Extensions** (GUI Agent, Knowledge \u0026 GraphRAG, AgentKit Integration).\n\n## Core Features\n\n### Core Framework\n- **Agent Core**: Agent execution engine based on 12-Factor Agents methodology, with Meta-Agent CEO dispatcher, agent team management, think-act loop, event-driven architecture, self-repair, and overflow recovery\n- **Embeddable ReActAgent (SDK primitive)**: Canonical async function-calling ReAct loop (`ainvoke`/`astream`) with a typed `AgentEvent` stream, multi-turn history in/out, parallel tool execution, and optional loop-detector / compactor / offloader injection — zero Studio/CLI coupling (legacy text-JSON `TextReActAgent` facade kept for compatibility)\n- **Unified Offload**: `Offloader` protocol + filesystem-backed `FileOffloader` keep large tool results / compressed context out of live history (inline reference placeholders, retrieved on demand); plus an in-workspace MCP gateway that runs MCP servers inside the sandbox\n- **Orchestration Engine**: Graph-based workflow engine + Flow system with decorators, execution plans, conditional routing, and parallel execution\n- **Tool System**: Unified tool interface with function decorators, MCP Hub (multi-server aggregation), remote tools v2, OpenAPI toolset, sandbox tools, skill bundles, and document routers\n- **Memory System**: Hierarchical memory (core / episodic / semantic), Mem0 deep integration, workspace memory, short-term memory, memory decay, hybrid search, compaction flush, MCP memory, and memory intelligence engine\n- **LLM Providers**: 15+ providers — OpenAI, Anthropic, Ollama, Gemini, Kimi/Moonshot, MiniMax, Ark/VolcEngine, Zhipu, Qianfan, Bailian/Dashscope — with response caching, transcript sanitizer, and failover routing\n- **Communication Protocols**: A2A inter-agent protocol (client / server / AgentCard / skill-as-tool), MCP resource access protocol\n- **Task Validation**: Pydantic-based output parsing, auto-repair, and guiderails\n\n### Avatar \u0026 Team Collaboration\n- **Avatar System**: Avatar registry (CRUD), group chat with multiple routing strategies (user-directed / meta-routed / round-robin)\n- **Meta-Agent Runtime**: CEO dispatcher with dynamic sub-agent orchestration, team management with concurrency limits, archived snapshots, and session isolation\n- **Collaboration Patterns**: Delegation, role-playing, conversation management, task locks, and collaboration metrics\n\n### Knowledge \u0026 Retrieval\n- **Knowledge Base**: Document processing pipeline with chunkers, readers, extractors, and graph builders (GraphRAG)\n- **Multi-Brain Knowledge**: Isolatable, per-avatar/session mountable \"doc brain + code brain\" architecture with cross-brain aggregated search\n- **Code Semantic Index**: Hybrid (vector + BM25) retrieval across multiple codebases, wired into the \"code brain\"\n- **Retrieval System**: Vector retriever, BM25 retriever, graph retriever, hybrid retriever, auto-retriever, and reranker\n- **Embeddings**: OpenAI, Bailian, SiliconFlow, LiteLLM, with smart routing\n\n### Skills \u0026 Self-Evolution\n- **Skill System**: Registration and full lifecycle management — dangerous-pattern security scan gate, 5-strategy fuzzy patch, `.changelog` versioning, source tagging, and per-skill enable/disable\n- **Skill Self-Evolution**: Captures tool-call observations at runtime and auto-distills new skills via background LLM session review; quality gate, usage stats, and deprecation form the lifecycle loop\n- **Extension Ecosystem (AGX Bundle)**: Bundle definitions (skills / mcp_servers / avatars / memory_templates), local install/uninstall, multi-source registry aggregated search\n\n### Long-Horizon Autonomous Coding\n- **Long-Run Orchestration**: Polls multiple task sources (manual queue / Cron / Linear / project features), per-task isolated workspaces, stall self-healing, continuation/failure dual-track backoff, incremental token accounting\n- **Project State Machine**: Disk-backed single source of truth with a versioned feature state machine, file locks / atomic writes, powering an auditable \"init → implement → verify → commit\" loop\n\n### Developer Experience\n- **CLI Tools** (`agx`): serve, studio, loop, run, project, deploy, codegen, docs, skills, hooks, debug, scaffold, and config management\n- **Web UI (Studio)**: FastAPI-based management server with session management, real-time WebSocket, and protocol support\n- **Desktop App**: Electron + React + Zustand + Vite, Pro/Lite dual mode (multi-pane / single-pane), command palette, settings panel, avatar sidebar, sub-agent panel, session history, and workspace panel\n- **IM Remote Gateway**: Remote command relay and reply delivery for Feishu / WeCom / DingTalk / personal WeChat (iLink) → cloud → local Agent\n- **Claude Code Bridge**: Token-protected local HTTP / NDJSON control plane driving local Claude Code in headless (stream-json) or visible TUI (PTY) modes\n\n### Enterprise Security\n- **Safety Layer**: Leak detection, input sanitizer, advanced injection detector, policy engine (rules / severity / actions), input validator, sandbox policy, and audit logging\n- **Sandbox**: Docker / Microsandbox / Subprocess / remote HTTP backends (tiered factory auto-selection); Jupyter kernel manager, stateful code interpreter, sandbox templates, JSONL execution audit\n- **Session Security**: Database-backed sessions, write locks, in-memory sessions, session-level multi-tenant (tenant_id) isolation\n\n### Observability \u0026 Evaluation\n- **Monitoring**: Complete callback system, real-time metrics, Prometheus/OpenTelemetry integration, trajectory analysis, span tree, WebSocket streaming\n- **Evaluation Framework**: EvalSet-based evaluation, LLM judge, composite judge, span evaluator, trajectory matcher, trace-to-evalset converter\n- **Data Export**: Multi-format export (JSON / CSV / Prometheus), time series analysis\n\n### Storage Layer\n- **Key-Value**: SQLite, Redis, PostgreSQL, MongoDB, InMemory\n- **Vector**: Milvus, Qdrant, Chroma, Faiss, PgVector, Pinecone, Weaviate\n- **Graph**: Neo4j, Nebula\n- **Object**: S3, GCS, Azure\n- **Unified Manager**: Storage router, migration support, unified storage interface\n\n### GUI Agent / Embodiment\n- **Action Reflection**: A/B/C result classification with heuristic and VLM reflection modes\n- **Stuck Detection \u0026 Recovery**: Consecutive failure detection, repeat pattern recognition, intelligent recovery strategy recommendation\n- **Action Caching**: Action-tree-based trajectory caching with exact and fuzzy matching (up to 9x speedup)\n- **REACT Output Parsing**: Standardized REACT format parsing with compact action schema\n- **Device-Cloud Routing**: Dynamic selection of on-device or cloud model based on task complexity and sensitivity\n- **DAG Task Verification**: DAG-based multi-path task verification with dual semantic dependencies\n- **Human-in-the-Loop**: Collector, component, and event model for human oversight\n\n## Quick Start\n\n### Installation\n\n#### Option 1: Install from PyPI (Recommended)\n\n```bash\n# Core install (lightweight, no torch, installs in seconds)\npip install agenticx\n\n# Install optional features as needed\npip install \"agenticx[memory]\"      # Memory: mem0, chromadb, qdrant, redis, milvus\npip install \"agenticx[document]\"    # Document processing: PDF, Word, PPT parsing\npip install \"agenticx[graph]\"       # Knowledge graph: networkx, neo4j, community detection\npip install \"agenticx[llm]\"         # Extra LLMs: anthropic, ollama\npip install \"agenticx[monitoring]\"  # Observability: prometheus, opentelemetry\npip install \"agenticx[mcp]\"         # MCP protocol\npip install \"agenticx[database]\"    # Database backends: postgres, SQLAlchemy\npip install \"agenticx[data]\"        # Data analysis: pandas, scikit-learn, matplotlib\npip install \"agenticx[ocr]\"         # OCR (pulls in torch ~2GB): easyocr\npip install \"agenticx[volcengine]\"  # Volcengine AgentKit\npip install \"agenticx[all]\"         # Everything\n```\n\n\u003e **Tip**: The core package includes only ~27 lightweight dependencies and installs in seconds. Heavy dependencies (torch, pandas, etc.) are optional extras - install only what you need.\n\n\u003e **Browser automation**: To run [browser-use](https://github.com/browser-use/browser-use) as an MCP server from AgenticX (`mcp_connect` / `mcp_call`), see [examples/browser-use-mcp.md](examples/browser-use-mcp.md).\n\n\u003e **Desktop MCP upgrades (2026-04)**: Near Settings now supports MCP brand auto-discovery (Cursor / Trae / Claude / OpenClaw / Hermes / Codex), built-in Monaco JSON editor with schema validation, and one-click install from ModelScope MCP marketplace.\n\n#### Option 2: Install from Source (Development)\n\n```bash\n# Clone repository\ngit clone https://github.com/DemonDamon/AgenticX.git\ncd AgenticX\n\n# Using uv (recommended, 10-100x faster than pip)\npip install uv\nuv pip install -e .                  # Core install\nuv pip install -e \".[memory,graph]\"  # Add optional features\nuv pip install -e \".[all]\"           # Everything\nuv pip install -e \".[dev]\"           # Development tools\n\n# Or using pip\npip install -e .\npip install -e \".[all]\"\n```\n\n#### Environment Setup\n\n```bash\n# Set environment variables\nexport OPENAI_API_KEY=\"your-api-key\"\nexport ANTHROPIC_API_KEY=\"your-api-key\"  # Optional\n```\n\n\u003e **Complete Installation Guide**: For system dependencies (antiword, tesseract) and advanced document processing features, see [INSTALL.md](INSTALL.md)\n\n### CLI Quick Start\n\nAfter installation, the `agx` command-line tool is available:\n\n```bash\n# View version\nagx --version\n\n# Create a new project\nagx project create my-agent --template basic\n\n# Start the API server\nagx serve --port 8000\n\n# Parse documents (PDF/PPT/Word etc.)\nagx mineru parse report.pdf --output ./parsed\n```\n\n\u003e **Full CLI Reference**: See [docs/cli.md](docs/cli.md) for complete command documentation.\n\n### Create Your First Agent\n\n```python\nfrom agenticx import Agent, Task, AgentExecutor\nfrom agenticx.llms import OpenAIProvider\n\n# Create agent\nagent = Agent(\n    id=\"data-analyst\",\n    name=\"Data Analyst\",\n    role=\"Data Analysis Expert\", \n    goal=\"Help users analyze and understand data\",\n    organization_id=\"my-org\"\n)\n\n# Create task\ntask = Task(\n    id=\"analysis-task\",\n    description=\"Analyze sales data trends\",\n    expected_output=\"Detailed analysis report\"\n)\n\n# Configure LLM\nllm = OpenAIProvider(model=\"gpt-4\")\n\n# Execute task\nexecutor = AgentExecutor(agent=agent, llm=llm)\nresult = executor.run(task)\nprint(result)\n```\n\n### Tool Usage Example\n\n```python\nfrom agenticx.tools import tool\n\n@tool\ndef calculate_sum(x: int, y: int) -\u003e int:\n    \"\"\"Calculate the sum of two numbers\"\"\"\n    return x + y\n\n@tool  \ndef search_web(query: str) -\u003e str:\n    \"\"\"Search web information\"\"\"\n    return f\"Search results: {query}\"\n\n# Agents will automatically invoke these tools\n```\n\n## Complete Examples\n\nWe provide rich examples demonstrating various framework capabilities:\n\n### Agent Core (M5)\n\n**Single Agent Example**\n```bash\n# Basic agent usage\npython examples/m5_agent_demo.py\n```\n- Demonstrates basic agent creation and execution\n- Tool invocation and error handling\n- Event-driven execution flow\n\n**Multi-Agent Collaboration**\n```bash\n# Multi-agent collaboration example\npython examples/m5_multi_agent_demo.py\n```\n- Multi-agent collaboration patterns\n- Task distribution and result aggregation\n- Inter-agent communication\n\n### Orchestration \u0026 Validation (M6 \u0026 M7)\n\n**Simple Workflow**\n```bash\n# Basic workflow orchestration\npython examples/m6_m7_simple_demo.py\n```\n- Workflow creation and execution\n- Task output parsing and validation\n- Conditional routing and error handling\n\n**Complex Workflow**\n```bash\n# Complex workflow orchestration\npython examples/m6_m7_comprehensive_demo.py\n```\n- Complex workflow graph structures\n- Parallel execution and conditional branching\n- Complete lifecycle management\n\n### Agent Communication (M8)\n\n**A2A Protocol Demo**\n```bash\n# Inter-agent communication protocol\npython examples/m8_a2a_demo.py\n```\n- Agent-to-Agent communication protocol\n- Distributed agent systems\n- Service discovery and skill invocation\n\n### Observability Monitoring (M9)\n\n**Complete Monitoring Demo**\n```bash\n# Observability module demo\npython examples/m9_observability_demo.py\n```\n- Real-time performance monitoring\n- Execution trajectory analysis\n- Failure analysis and recovery recommendations\n- Data export and report generation\n\n### Memory System\n\n**Basic Memory Usage**\n```bash\n# Memory system example\npython examples/memory_example.py\n```\n- Long-term memory storage and retrieval\n- Context memory management\n\n**Healthcare Scenario**\n```bash\n# Healthcare memory scenario\npython examples/mem0_healthcare_example.py  \n```\n- Medical knowledge memory and application\n- Personalized patient information management\n\n### Human-in-the-Loop\n\n**Human Intervention Flow**\n```bash\n# Human-in-the-loop example\npython examples/human_in_the_loop_example.py\n```\n- Human approval workflows\n- Human-machine collaboration patterns\n- Risk control mechanisms\n\nDetailed documentation: [examples/README_HITL.md](examples/README_HITL.md)\n\n### LLM Integration\n\n**Chatbot**\n```bash\n# LLM chat example\npython examples/llm_chat_example.py\n```\n- Multi-model support demonstration\n- Streaming response handling\n- Cost control and monitoring\n\n### Security Sandbox\n\n**Code Execution Sandbox**\n```bash\n# Micro-sandbox example\npython examples/microsandbox_example.py\n```\n- Secure code execution environment\n- Resource limits and isolation\n\nTechnical blog: [examples/microsandbox_blog.md](examples/microsandbox_blog.md)\n\n### Intent Recognition Service\n\n**Intelligent Intent Recognition System**\n```bash\n# Intent recognition service example\npython examples/agenticx-for-intent-recognition/main.py\n```\n\nA production-grade, layered intent recognition service built entirely on the AgenticX framework, demonstrating real-world usage of Agents, Workflows, Tools, and Storage systems.\n\nArchitecture:\n- **Agent Layer**: Hierarchical agent design — a base `IntentRecognitionAgent` (LLM-powered) with specialized agents (`GeneralIntentAgent`, `SearchIntentAgent`, `FunctionIntentAgent`) for fine-grained classification\n- **Workflow Engine**: Pipeline-based orchestration — preprocessing → intent classification → entity extraction → rule matching → post-processing; plus dedicated workflows for each intent type\n- **Tool System**: Hybrid entity extraction (`UIE` + `LLM` + `Rule` extractors with confidence-weighted fusion), regex/full-text matching, and a full post-processing suite (confidence adjustment, conflict resolution, entity optimization, intent refinement)\n- **API Gateway**: Async service layer with rate limiting, concurrent control, batch processing, health checks, and performance metrics\n- **Storage**: SQLite-backed data persistence for training data management via `UnifiedStorageManager`\n- **Data Models**: Pydantic-based type-safe data contracts for API requests/responses and domain objects\n\nKey capabilities:\n- **Three-tier Intent Classification**: General dialogue (greetings, chitchat), information search (factual/how-to/comparison queries), and function/tool invocation\n- **Hybrid Entity Extraction**: Combines UIE models, LLM, and rule-based extractors with intelligent fusion strategies\n- **Full Post-processing Pipeline**: Confidence adjustment, conflict resolution, entity optimization, and intent refinement\n- **Extensible Design**: Add new intent types by simply creating a new agent and workflow — zero changes to existing code\n\nSee: [examples/agenticx-for-intent-recognition/](examples/agenticx-for-intent-recognition/)\n\n### GUI Agent / Embodiment (M16)\n\n**GUI Automation Agent**\n```bash\n# GUI Agent example\npython examples/agenticx-for-guiagent/AgenticX-GUIAgent/main.py\n```\n- Complete GUI automation framework with human-aligned learning\n- Action reflection (A/B/C classification) and stuck detection\n- Action caching system for performance optimization\n- REACT output parsing and compact action schema\n- Device-Cloud routing for intelligent model selection\n- DAG-based task verification\n\nKey capabilities:\n- **Action Reflection**: Automatic action result classification (success/wrong_state/no_change)\n- **Stuck Detection**: Continuous failure detection and recovery strategy recommendation\n- **Action Caching**: Trajectory caching with exact and fuzzy matching (up to 9x speedup)\n- **REACT Parsing**: Standardized REACT format output parsing\n- **Smart Routing**: Dynamic device-cloud model selection based on task complexity and sensitivity\n- **DAG Verification**: Multi-path task verification with dual-semantic dependencies\n\nSee: [examples/agenticx-for-guiagent/](examples/agenticx-for-guiagent/)\n\n### More Application Examples\n\n| Project | Description | Path |\n|---------|-------------|------|\n| **Agent Skills** | Skill discovery, matching, and SOP-driven skill execution for agents | [examples/agenticx-for-agent-skills/](examples/agenticx-for-agent-skills/) |\n| **AgentKit** | Volcengine AgentKit integration with Docker-ready agent deployment | [examples/agenticx-for-agentkit/](examples/agenticx-for-agentkit/) |\n| **ChatBI** | Conversational BI — natural language to data insights | [examples/agenticx-for-chatbi/](examples/agenticx-for-chatbi/) |\n| **Deep Research** | Multi-source deep research and report generation | [examples/agenticx-for-deepresearch/](examples/agenticx-for-deepresearch/) |\n| **Doc Parser** | Intelligent document parsing (PDF, Word, PPT) | [examples/agenticx-for-docparser/](examples/agenticx-for-docparser/) |\n| **Finance** | Financial news hunting and analysis | [examples/agenticx-for-finance/](examples/agenticx-for-finance/) |\n| **Future Prediction** | Predictive analysis and forecasting | [examples/agenticx-for-future-prediction/](examples/agenticx-for-future-prediction/) |\n| **GraphRAG** | Knowledge graph-enhanced retrieval-augmented generation | [examples/agenticx-for-graphrag/](examples/agenticx-for-graphrag/) |\n| **Math Modeling** | Mathematical modeling assistant | [examples/agenticx-for-math-modeling/](examples/agenticx-for-math-modeling/) |\n| **Model Architecture Discovery** | Automated model architecture search and discovery | [examples/agenticx-for-modelarch-discovery/](examples/agenticx-for-modelarch-discovery/) |\n| **Query Optimizer** | SQL/query optimization agent | [examples/agenticx-for-queryoptimizer/](examples/agenticx-for-queryoptimizer/) |\n| **Sandbox** | Secure code execution sandbox | [examples/agenticx-for-sandbox/](examples/agenticx-for-sandbox/) |\n| **Spec Coding** | Specification-driven code generation | [examples/agenticx-for-spec-coding/](examples/agenticx-for-spec-coding/) |\n| **Vibe Coding** | AI-assisted creative/vibe coding | [examples/agenticx-for-vibecoding/](examples/agenticx-for-vibecoding/) |\n\n## Technical Architecture\n\n```mermaid\ngraph TD\n    subgraph \"User Interface Layer\"\n        Desktop[\"Desktop App (Electron + React)\"]\n        CLI[\"CLI (agx serve / loop / run / project)\"]\n        SDK[Python SDK]\n    end\n\n    subgraph \"Studio Runtime Layer\"\n        StudioServer[\"Studio Server (FastAPI)\"]\n        SessionMgr[Session Manager]\n        MetaAgent[\"Meta-Agent (CEO Dispatcher)\"]\n        TeamMgr[Agent Team Manager]\n        AvatarSys[\"Avatar \u0026 Group Chat\"]\n    end\n\n    subgraph \"Core Framework Layer\"\n        subgraph \"Orchestration\"\n            WorkflowEngine[Workflow Engine]\n            Flow[\"Flow System\"]\n        end\n        subgraph \"Execution\"\n            AgentRuntime[\"Agent Runtime (Studio)\"]\n            AgentExecutor[\"Agent Executor (Core)\"]\n            TaskValidator[Task Validator \u0026 Output Parser]\n        end\n        subgraph \"Core Components\"\n            Agent[Agent]\n            Task[Task]\n            Tool[Tool System \u0026 MCP Hub]\n            Memory[\"Memory (Mem0 / Short-term / Workspace)\"]\n            LLM[\"LLM Providers (OpenAI / Anthropic / Ollama / 10+)\"]\n        end\n        Collaboration[\"Collaboration \u0026 Delegation\"]\n        Hooks[\"Hooks System\"]\n    end\n\n    subgraph \"Platform Services Layer\"\n        subgraph \"Observability\"\n            Monitoring[\"Monitoring \u0026 Trajectory\"]\n            Prometheus[Prometheus / OpenTelemetry]\n        end\n        subgraph \"Protocols\"\n            A2A[\"A2A Protocol\"]\n            MCP[\"MCP Protocol\"]\n        end\n        subgraph \"Security\"\n            Safety[\"Safety Layer (Leak Detection / Sanitizer / Policy)\"]\n            Sandbox[\"Execution Sandbox\"]\n        end\n        subgraph \"Storage\"\n            KVStore[\"Key-Value (SQLite / Redis)\"]\n            VectorStore[\"Vector (Milvus / Qdrant / Chroma)\"]\n            GraphStore[\"Graph (Neo4j / NetworkX)\"]\n        end\n    end\n\n    subgraph \"Domain Extensions\"\n        Embodiment[\"GUI Agent / Embodiment\"]\n        Knowledge[\"Knowledge \u0026 GraphRAG\"]\n    end\n\n    Desktop --\u003e StudioServer\n    CLI --\u003e StudioServer\n    SDK --\u003e AgentExecutor\n\n    StudioServer --\u003e SessionMgr\n    SessionMgr --\u003e MetaAgent\n    MetaAgent --\u003e TeamMgr\n    MetaAgent --\u003e AvatarSys\n    TeamMgr --\u003e AgentRuntime\n\n    AgentRuntime --\u003e Agent\n    AgentExecutor --\u003e Agent\n    WorkflowEngine --\u003e AgentExecutor\n\n    Agent --\u003e Tool\n    Agent --\u003e Memory\n    Agent --\u003e LLM\n    Agent --\u003e Hooks\n\n    AgentRuntime --\u003e Monitoring\n    AgentExecutor --\u003e Monitoring\n    Agent --\u003e A2A\n    Tool --\u003e MCP\n\n    Agent --\u003e Safety\n    Memory --\u003e KVStore\n    Memory --\u003e VectorStore\n    Knowledge --\u003e GraphStore\n```\n\n## Development Progress\n\n### ✅ Completed Modules (M1-M11, M13-M17)\n\n| Module | Status | Description |\n|---------|--------|-------------|\n| **M1** | ✅ | Core Abstraction Layer — Agent, Task, Tool, Workflow, Event Bus, Component, and Pydantic data contracts |\n| **M2** | ✅ | LLM Service Layer — 15+ providers (OpenAI / Anthropic / Ollama / Gemini / Kimi / MiniMax / Ark / Zhipu / Qianfan / Bailian), response caching, failover routing |\n| **M3** | ✅ | Tool System — Function decorators, MCP Hub, remote tools v2, OpenAPI toolset, sandbox tools, skill bundles, document routers |\n| **M4** | ✅ | Memory System — Hierarchical (core / episodic / semantic), Mem0, workspace, short-term, memory decay, hybrid search, memory intelligence engine |\n| **M5** | ✅ | Agent Core — Meta-Agent CEO dispatcher, think-act loop, event-driven architecture, self-repair, overflow recovery, reflection |\n| **M6** | ✅ | Task Validation — Pydantic-based output parsing, auto-repair, guiderails |\n| **M7** | ✅ | Orchestration Engine — Graph-based workflow engine + Flow system with decorators, execution plans, conditional routing, parallel execution |\n| **M8** | ✅ | Communication Protocols — A2A (client / server / AgentCard / skill-as-tool), MCP resource access, AGUI protocol |\n| **M9** | ✅ | Observability — Callbacks, real-time monitoring, trajectory analysis, span tree, WebSocket streaming, Prometheus / OpenTelemetry integration |\n| **M10** | ✅ | Developer Experience — CLI (`agx` with 15+ commands), Studio Server (FastAPI), Desktop App (Electron + React + Zustand, Pro/Lite dual mode) |\n| **M11** | ✅ | Enterprise Security — Safety layer (leak detection / sanitizer / injection detector / policy / audit), Sandbox (Docker / Microsandbox / Subprocess / Jupyter kernel / code interpreter) |\n| **M13** | ✅ | Knowledge \u0026 Retrieval — Knowledge base with document processing, chunkers, graphers (GraphRAG), readers; retrieval (vector / BM25 / graph / hybrid / auto); embeddings (OpenAI / Bailian / SiliconFlow / LiteLLM) |\n| **M14** | ✅ | Avatar \u0026 Collaboration — Avatar registry, group chat (user-directed / meta-routed / round-robin), delegation, role-playing, conversation patterns, team management |\n| **M15** | ✅ | Evaluation Framework — EvalSet, LLM judge, composite judge, span evaluator, trajectory matcher, trace converter |\n| **M16** | ✅ | Embodiment — GUI Agent framework with action reflection, stuck detection, action caching, REACT parsing, device-cloud routing, DAG verification, human-in-the-loop |\n| **M17** | ✅ | Storage Layer — Key-Value (SQLite / Redis / PostgreSQL / MongoDB), Vector (Milvus / Qdrant / Chroma / Faiss / PgVector / Pinecone / Weaviate), Graph (Neo4j / Nebula), Object (S3 / GCS / Azure) |\n\n### 🚧 Planned Modules\n\n| Module | Status | Description |\n|---------|--------|-------------|\n| **M12** | 🚧 | Agent Evolution — Architecture search, knowledge distillation, adaptive planning |\n| **M18** | 🚧 | Multi-tenancy \u0026 RBAC — Session-level `tenant_id` isolation landed; fine-grained permission control in progress |\n\n### 🆕 Recent Capability Additions (H1 2026)\n\n| Capability | Status | Description |\n|------------|--------|-------------|\n| **Skill Self-Evolution** | ✅ | Runtime tool-call observation capture, auto-skill creation via session review, quality gate / usage stats / deprecation loop (`learning`) |\n| **Multi-Brain Knowledge** | ✅ | Isolatable/mountable \"doc brain + code brain\" with cross-brain search (`brain`) + multi-codebase hybrid semantic index (`code_index`) |\n| **Long-Horizon Coding** | ✅ | Long-run orchestration (multi-source / isolated workspaces / stall self-healing / continuation backoff, `longrun`) + disk-backed project state machine (`project_state`) |\n| **IM Channel Integration** | ✅ | Remote command gateway for Feishu / WeCom / DingTalk / personal WeChat (iLink) (`gateway`) |\n| **Claude Code Bridge** | ✅ | Token-protected local HTTP/NDJSON control plane, headless / visible TUI dual mode (`cc_bridge`) |\n| **Extension Ecosystem** | ✅ | AGX Bundle definitions, local install/uninstall, multi-source registry aggregated search (`extensions`) |\n| **Embeddable ReActAgent** | ✅ | Canonical async function-calling ReAct SDK primitive with typed event stream, multi-turn history, parallel tools, optional loop-detector / compactor / offloader, zero Studio/CLI coupling (`agents`) |\n| **Unified Offload \u0026 MCP Gateway** | ✅ | `Offloader` protocol + `FileOffloader` for out-of-history large payloads, plus in-workspace MCP gateway (AgentScope v2 P0 internalization, `core.offload` / `sandbox.mcp_gateway`) |\n\n## Core Advantages\n\n- **Unified Abstraction**: Clear and consistent core abstractions, avoiding conceptual confusion\n- **Pluggable Architecture**: All components are replaceable, avoiding vendor lock-in\n- **Enterprise-Grade Monitoring**: Complete observability, production-ready\n- **Security First**: Built-in security mechanisms and multi-tenant support\n- **High Performance**: Optimized execution engine and concurrent processing\n- **Rich Ecosystem**: Complete toolset and example library\n\n## System Requirements\n\n- **Python**: 3.10+\n- **Memory**: 4GB+ RAM recommended\n- **System**: Windows / Linux / macOS\n- **Core Dependencies**: ~27 lightweight packages, installs in seconds (see `pyproject.toml`)\n- **Optional Dependencies**: 15 feature groups available via `pip install \"agenticx[xxx]\"`\n\n## Contributing\n\nWe welcome community contributions! Please refer to:\n\n1. Submit Issues to report bugs or request features\n2. Fork the project and create feature branches\n3. Submit Pull Requests, ensuring all tests pass\n4. Participate in code reviews and discussions\n\n## Acknowledgements / Upstream Credits\n\nThe personal WeChat (iLink) channel integration in AgenticX was built on top of the **openilink-sdk-go** library from [OpeniLink Hub](https://github.com/openilink/openilink-hub). We specifically relied on:\n\n- **QR code binding flow** — `FetchQRCode` / `PollQRStatus` APIs for the scan-to-bind UX\n- **Message monitoring** — `client.Monitor()` for real-time inbound message streaming  \n- **Outbound messaging** — `SendText` / `Push` for reply delivery with `context_token` routing\n- **CDN media handling** — `DownloadMedia` / `DownloadVoice` for encrypted WeChat media\n\nOpeniLink Hub's [OpenClaw App](https://github.com/openilink/openilink-hub) also demonstrated an AI Agent gateway integration pattern that informed our adapter architecture.\n\nWe did **not** include OpeniLink Hub's web console, App Marketplace, or multi-bot management features. AgenticX's **core multi-agent runtime**, **session management**, and **Desktop UI** remain fully independent implementations.\n\n\u003e OpeniLink Hub — MIT License — [github.com/openilink/openilink-hub](https://github.com/openilink/openilink-hub)\n\nAdditional reference: [WorkBuddy — WeixinBot Guide](https://www.codebuddy.cn/docs/workbuddy/WeixinBot-Guide) for iLink protocol usage patterns.\n\n**Desktop development:** The iLink Go sidecar binary is **not** committed to this repository. Before using the personal WeChat bridge locally, run `make build` in [`packaging/wechat-sidecar/`](packaging/wechat-sidecar/) (requires Go 1.22+). See [`packaging/wechat-sidecar/README.md`](packaging/wechat-sidecar/README.md).\n\n## License\n\nThis project is licensed under the Apache License, Version 2.0 — see the [LICENSE](LICENSE) file for details.\n\n## Star History\n\n[![Star History Chart](https://api.star-history.com/svg?repos=DemonDamon/AgenticX\u0026type=Date)](https://star-history.com/#DemonDamon/AgenticX\u0026Date)\n\n## Acknowledgments\n\nAgenticX would not exist in its current form without the inspiration, architectural ideas, and engineering wisdom we drew from the open-source community. We have studied the following projects in depth, and we are genuinely grateful to every author, contributor, and community behind them.\n\n| Project | Repository | What we learned |\n|---------|-----------|-----------------|\n| **A2A** | [a2aproject/A2A](https://github.com/a2aproject/A2A) | Agent-to-Agent protocol design |\n| **AgentCPM-GUI** | [OpenBMB/AgentCPM-GUI](https://github.com/OpenBMB/AgentCPM-GUI) | Compact GUI action schema \u0026 RFT training |\n| **ADK Python** | [google/adk-python](https://github.com/google/adk-python) | Agent lifecycle, runner abstractions |\n| **ag-ui** | [ag-ui-protocol/ag-ui](https://github.com/ag-ui-protocol/ag-ui) | Agent–UI streaming protocol |\n| **AgentKit SDK** | [volcengine/agentkit-sdk-python](https://github.com/volcengine/agentkit-sdk-python) | Agent deployment \u0026 skill packaging |\n| **AgentRun SDK** | [Serverless-Devs/agentrun-sdk-python](https://github.com/Serverless-Devs/agentrun-sdk-python) | Serverless agent runtime patterns |\n| **AgentScope** | [agentscope-ai/agentscope](https://github.com/agentscope-ai/agentscope) | Multi-agent communication \u0026 pipeline |\n| **Agno** | [agno-agi/agno](https://github.com/agno-agi/agno) | Lightweight agent framework design |\n| **Camel** | [camel-ai/camel](https://github.com/camel-ai/camel) | Role-playing agents \u0026 society simulation |\n| **Cherry Studio** | [CherryHQ/cherry-studio](https://github.com/CherryHQ/cherry-studio) | Desktop UX, MCP integration, skill system |\n| **Claude Code** | [anthropics/claude-code](https://github.com/anthropics/claude-code) | Agentic CLI UX \u0026 plugin architecture |\n| **CLI-Anything** | [HKUDS/CLI-Anything](https://github.com/HKUDS/CLI-Anything) | CLI-native agent harness |\n| **ClawTeam** | [HKUDS/ClawTeam](https://github.com/HKUDS/ClawTeam) | Multi-agent team coordination |\n| **CodexMonitor** | [Dimillian/CodexMonitor](https://github.com/Dimillian/CodexMonitor) | Desktop monitoring \u0026 Tauri app patterns |\n| **CrewAI** | [crewAIInc/crewAI](https://github.com/crewAIInc/crewAI) | Crew orchestration, flow \u0026 memory system |\n| **DeepWiki Open** | [AsyncFuncAI/deepwiki-open](https://github.com/AsyncFuncAI/deepwiki-open) | Repository-level knowledge indexing |\n| **Deer Flow** | [bytedance/deer-flow](https://github.com/bytedance/deer-flow) | Deep research workflow \u0026 skill harness |\n| **Eigent** | [eigent-ai/eigent](https://github.com/eigent-ai/eigent) | Multi-agent workforce \u0026 SSE event spec |\n| **Iron Claw** | [nearai/ironclaw](https://github.com/nearai/ironclaw) | Agent evaluation \u0026 benchmark harness |\n| **JoyAgent / JD Genie** | [jd-opensource/joyagent-jdgenie](https://github.com/jd-opensource/joyagent-jdgenie) | Enterprise agent orchestration |\n| **Khazix Skills** | [KKKKhazix/Khazix-Skills](https://github.com/KKKKhazix/Khazix-Skills) | Skill module structure \u0026 packaging |\n| **Lobe Icons** | [lobehub/lobe-icons](https://github.com/lobehub/lobe-icons) | AI provider icon design system |\n| **LoongSuite Python Agent** | [alibaba/loongsuite-python-agent](https://github.com/alibaba/loongsuite-python-agent) | OpenTelemetry GenAI instrumentation |\n| **MAI-UI** | [Tongyi-MAI/MAI-UI](https://github.com/Tongyi-MAI/MAI-UI) | Device-cloud collaboration \u0026 GUI grounding |\n| **Microsandbox** | [zerocore-ai/microsandbox](https://github.com/zerocore-ai/microsandbox) | Lightweight sandboxed code execution |\n| **MobiAgent** | [IPADS-SAI/MobiAgent](https://github.com/IPADS-SAI/MobiAgent) | Mobile multi-stage planning |\n| **MobileAgent** | [X-PLUG/MobileAgent](https://github.com/X-PLUG/MobileAgent) | Multi-agent mobile GUI automation |\n| **Model Context Protocol** | [modelcontextprotocol/modelcontextprotocol](https://github.com/modelcontextprotocol/modelcontextprotocol) | Standardized LLM tool/resource protocol |\n| **NVIDIA NemoClaw** | [NVIDIA/NemoClaw](https://github.com/NVIDIA/NemoClaw) | GPU-accelerated agent plugin system |\n| **OpenClaw** | [openclaw/openclaw](https://github.com/openclaw/openclaw) | Open desktop agent platform \u0026 extensions |\n| **OpenSandbox** | [alibaba/OpenSandbox](https://github.com/alibaba/OpenSandbox) | Container-based code sandbox |\n| **OpenShell** | [NVIDIA/OpenShell](https://github.com/NVIDIA/OpenShell) | Rust-based secure agent shell |\n| **OpenSkills** | [numman-ali/openskills](https://github.com/numman-ali/openskills) | Skill registry \u0026 discovery |\n| **OWL** | [camel-ai/owl](https://github.com/camel-ai/owl) | Embodied multi-agent collaboration |\n| **Pydantic AI** | [pydantic/pydantic-ai](https://github.com/pydantic/pydantic-ai) | Type-safe agent \u0026 eval framework |\n| **Refly** | [refly-ai/refly](https://github.com/refly-ai/refly) | AI-native knowledge canvas UX |\n| **Serverless Devs** | [Serverless-Devs/Serverless-Devs](https://github.com/Serverless-Devs/Serverless-Devs) | Serverless agent deployment toolchain |\n| **Skills** | [anthropics/skills](https://github.com/anthropics/skills) | Skill definition format \u0026 lifecycle |\n| **Spring AI** | [spring-projects/spring-ai](https://github.com/spring-projects/spring-ai) | Enterprise AI abstraction patterns |\n| **SWE-agent** | [SWE-agent/SWE-agent](https://github.com/SWE-agent/SWE-agent) | Software engineering agent \u0026 ACR loop |\n| **VE ADK** | [volcengine/veadk-python](https://github.com/volcengine/veadk-python) | Skills system \u0026 cloud-native A2A |\n| **ZeroBoot** | [zerobootdev/zeroboot](https://github.com/zerobootdev/zeroboot) | Zero-config agent bootstrapping |\n\nThank you for building in the open. Your work has been a constant source of insight and motivation for the AgenticX team.\n\n---\n\n\u003cdiv align=\"center\"\u003e\n\n**If AgenticX helps you, please give us a Star!**\n\n[GitHub](https://github.com/DemonDamon/AgenticX) • [Documentation](coming-soon) • [Examples](examples/) • [Discussions](https://github.com/DemonDamon/AgenticX/discussions)\n\n\u003c/div\u003e","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2FDemonDamon%2FAgenticX","html_url":"https://awesome.ecosyste.ms/projects/github.com%2FDemonDamon%2FAgenticX","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2FDemonDamon%2FAgenticX/lists"}