[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-72366":3},{"id":4,"name":5,"fullName":6,"owner":5,"repo":5,"description":7,"homepage":8,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":14,"contributorsCount":15,"subscribersCount":15,"size":15,"stars1d":16,"stars7d":17,"stars30d":18,"stars90d":15,"forks30d":15,"starsTrendScore":19,"compositeScore":20,"rankGlobal":9,"rankLanguage":9,"license":21,"archived":22,"fork":22,"defaultBranch":23,"hasWiki":22,"hasPages":24,"topics":25,"createdAt":9,"pushedAt":9,"updatedAt":37,"readmeContent":38,"aiSummary":39,"trendingCount":15,"starSnapshotCount":15,"syncStatus":40,"lastSyncTime":41,"discoverSource":42},72366,"EvoAgentX","EvoAgentX\u002FEvoAgentX","🚀 EvoAgentX: Building a Self-Evolving Ecosystem of AI Agents","https:\u002F\u002Fevoagentx.github.io\u002FEvoAgentX\u002F",null,"Python",3066,272,23,13,0,7,15,83,21,29.31,"Other",false,"main",true,[26,27,28,29,30,31,32,33,34,35,36],"agent","ai","ai-agents","llms","memory","multi-agent-systems","natural-language-processing","rag","self-evolving","tool","tools","2026-06-12 02:03:02","\u003C!-- Add logo here -->\n\u003Cdiv align=\"center\">\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX\">\n    \u003Cimg src=\".\u002Fassets\u002FEAXLoGo.svg\" alt=\"EvoAgentX\" width=\"50%\">\n  \u003C\u002Fa>\n\u003C\u002Fdiv>\n\n\u003Ch2 align=\"center\">\n    Building a Self-Evolving Ecosystem of AI Agents\n\u003C\u002Fh2>\n\n\u003Cdiv align=\"center\">\n\n[![EvoAgentX Homepage](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FEvoAgentX-Homepage-blue?logo=homebridge)](https:\u002F\u002Fevoagentx.org\u002F)\n[![Docs](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F-Documentation-0A66C2?logo=readthedocs&logoColor=white&color=7289DA&labelColor=grey)](https:\u002F\u002FEvoAgentX.github.io\u002FEvoAgentX\u002F)\n[![Discord](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FChat-Discord-5865F2?&logo=discord&logoColor=white)](https:\u002F\u002Fdiscord.gg\u002FXWBZUJFwKe)\n[![Twitter](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FFollow-@EvoAgentX-e3dee5?&logo=x&logoColor=white)](https:\u002F\u002Fx.com\u002FEvoAgentX)\n[![Wechat](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FWeChat-EvoAgentX-brightgreen?logo=wechat&logoColor=white)](.\u002Fassets\u002Fwechat_info.md)\n[![GitHub star chart](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002FEvoAgentX\u002FEvoAgentX?style=social)](https:\u002F\u002Fstar-history.com\u002F#EvoAgentX\u002FEvoAgentX)\n[![GitHub fork](https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fforks\u002FEvoAgentX\u002FEvoAgentX?style=social)](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX\u002Ffork)\n[![License](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-MIT-blue.svg?)](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX\u002Fblob\u002Fmain\u002FLICENSE)\n\u003C!-- [![EvoAgentX Homepage](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FEvoAgentX-Homepage-blue?logo=homebridge)](https:\u002F\u002FEvoAgentX.github.io\u002FEvoAgentX\u002F) -->\n\u003C!-- [![hf_space](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F%F0%9F%A4%97%20Hugging%20Face-EvoAgentX-ffc107?color=ffc107&logoColor=white)](https:\u002F\u002Fhuggingface.co\u002FEvoAgentX) -->\n\u003C\u002Fdiv>\n\n\u003Cdiv align=\"center\">\n\n\u003Ch3 align=\"center\">\n\n\u003Ca href=\".\u002FREADME.md\" style=\"text-decoration: underline;\">English\u003C\u002Fa> | \u003Ca href=\".\u002FREADME-zh.md\">简体中文\u003C\u002Fa>\n\n\u003C\u002Fh3>\n\n\u003C\u002Fdiv>\n\n\n\n## What is EvoAgentX\nEvoAgentX is an open-source framework for building, evaluating, and evolving LLM-based agents or agentic workflows in an automated, modular, and goal-driven manner. At its core, EvoAgentX enables developers and researchers to move beyond static prompt chaining or manual workflow orchestration. It introduces a self-evolving agent ecosystem, where AI agents can be constructed, assessed, and optimized through iterative feedback loops—much like how software is continuously tested and improved.\n\n### ✨ Key Features\n\n- 🧱 **Agent Workflow Autoconstruction**\n  \n  From a single prompt, EvoAgentX builds structured, multi-agent workflows tailored to the task.\n\n- 🔍 **Built-in Evaluation**\n  \n  It integrates automatic evaluators to score agent behavior using task-specific criteria.\n\n- 🔁 **Self-Evolution Engine**\n  \n  Agents don’t just work—they learn. EvoAgentX improves workflows using self-evolving algorithms.\n- 🧩 **Plug-and-Play Compatibility**\n  \n  Easily integrate original [OpenAI](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX\u002Fblob\u002Fmain\u002Fevoagentx\u002Fmodels\u002Fopenai_model.py) and [qwen](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX\u002Fblob\u002Fmain\u002Fevoagentx\u002Fmodels\u002Faliyun_model.py) or other popular models, including Claude, Deepseek, kimi models through ([LiteLLM](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX\u002Fblob\u002Fmain\u002Fevoagentx\u002Fmodels\u002Flitellm_model.py), [siliconflow](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX\u002Fblob\u002Fmain\u002Fevoagentx\u002Fmodels\u002Fsiliconflow_model.py) or [openrouter](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX\u002Fblob\u002Fmain\u002Fevoagentx\u002Fmodels\u002Fopenrouter_model.py)). If you want to use LLMs locally deployed on your own machine, you can try LiteLLM. \n\n- 🧰 **Comprehensive Built-in Tools**\n  \n  EvoAgentX ships with a rich set of built-in tools that empower agents to interact with real-world environments.\n\n- 🧠 **Memory Module**\n  \n  EvoAgentX supports both ephemeral (short-term) and persistent (long-term) memory systems.\n\n- 🧑‍💻 **Human-in-the-Loop (HITL) Interactions**\n  \n  EvoAgentX supports interactive workflows where humans review, correct, and guide agent behavior.\n\n\n### 🚀 What You Can Do with EvoAgentX\n\nEvoAgentX isn’t just a framework — it’s your **launchpad for real-world AI agents**.\n\nWhether you're an AI researcher, workflow engineer, or startup team, EvoAgentX helps you **go from a vague idea to a fully functional agentic system** — with minimal engineering and maximum flexibility.\n\nHere’s how:\n\n- 🔍 **Struggling to improve your workflows?**  \n  EvoAgentX can **automatically evolve and optimize your agentic workflows** using SOTA self-evolving algorithms, driven by your dataset and goals.\n- 🧑‍💻 **Want to supervise the agent and stay in control?**  \n  Insert yourself into the loop! EvoAgentX supports **Human-in-the-Loop (HITL)** checkpoints, so you can step in, review, or guide the workflow as needed — and step out again.\n\n- 🧠 **Frustrated by agents that forget everything?**  \n  EvoAgentX provides **both short-term and long-term memory modules**, enabling your agents to remember, reflect, and improve across interactions.\n\n- ⚙️ **Lost in manual workflow orchestration?**  \n  Just describe your goal — EvoAgentX will **automatically assemble a multi-agent workflow** that matches your intent.\n\n- 🌍 **Want your agents to actually *do* things?**  \n  With a rich library of built-in tools (search, code, browser, file I\u002FO, APIs, and more), EvoAgentX empowers agents to **interact with the real world**, not just talk about it.\n\n\n\n## 🔥 EAX Latest News\n\n- **[Aug 2025]** 🚀 **New Survey Released!**  \n  Our team just published a comprehensive survey on **Self-Evolving AI Agents**—exploring how agents can learn, adapt, and optimize over time.  \n  👉 [Read it on arXiv](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.07407)\n  👉 [Check the repo](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FAwesome-Self-Evolving-Agents)\n\n- **[July 2025]** 📚 **EvoAgentX Framework Paper is Live!**  \n  We officially published the EvoAgentX framework paper on arXiv, detailing our approach to building evolving agentic workflows.  \n  👉 [Check it out](https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.03616)\n\n- **[July 2025]** ⭐️ **1,000 Stars Reached!**  \n  Thanks to our amazing community, **EvoAgentX** has surpassed 1,000 GitHub stars!\n\n- **[May 2025]** 🚀 **Official Launch!**  \n  **EvoAgentX** is now live! Start building self-evolving AI workflows from day one.  \n  🔧 [Get Started on GitHub](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX)\n\n## ⚡ Get Started\n- [🔥 Latest News](#-latest-news)\n- [⚡ Get Started](#-get-started)\n- [Installation](#installation)\n- [LLM Configuration](#llm-configuration)\n  - [API Key Configuration](#api-key-configuration)\n  - [Configure and Use the LLM](#configure-and-use-the-llm)\n- [Automatic WorkFlow Generation](#automatic-workflow-generation)\n- [EvoAgentX Built-in Tools Summary](#-evoagentx-built-in-tools-summary)\n- [Tool-Enabled Workflows Generation](#tool-enabled-workflows-generation)\n- [Demo Video](#demo-video)\n  - [✨ Final Results](#-final-results)\n- [Evolution Algorithms](#evolution-algorithms)\n  - [📊 Results](#-results)\n- [Applications](#applications)\n- [Tutorial and Use Cases](#tutorial-and-use-cases)\n- [🗣️ EvoAgentX TALK](#evoagentx-talk)\n- [🎯 Roadmap](#-roadmap)  \n- [🙋 Support](#-support)\n  - [Join the Community](#join-the-community)\n  - [Add the meeting to your calendar](#add-the-meeting-to-your-calendar)\n  - [Contact Information](#contact-information)\n  - [Community Call](#community-call)\n- [🙌 Contributing to EvoAgentX](#-contributing-to-evoagentx)\n- [📖 Citation](#-citation)\n- [📚 Acknowledgements](#-acknowledgements)\n- [📄 License](#-license)\n\n\n\n## Installation\n\nWe recommend installing EvoAgentX using `pip`:\n\n```bash\npip install evoagentx\n```\nor install from source:\n\n```bash\npip install git+https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX.git\n```\n\nFor local development or detailed setup (e.g., using conda), refer to the [Installation Guide for EvoAgentX](.\u002Fdocs\u002Finstallation.md).\n\n\u003Cdetails>\n\u003Csummary>Example (optional, for local development):\u003C\u002Fsummary>\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX.git\ncd EvoAgentX\n# Create a new conda environment\nconda create -n evoagentx python=3.11\n\n# Activate the environment\nconda activate evoagentx\n\n# Install the package\npip install -r requirements.txt\n# OR install in development mode\npip install -e .\n```\n\u003C\u002Fdetails>\n\n## LLM Configuration\n\n### API Key Configuration \n\nTo use LLMs with EvoAgentX (e.g., OpenAI), you must set up your API key.\n\n\u003Cdetails>\n\u003Csummary>Option 1: Set API Key via Environment Variable\u003C\u002Fsummary> \n\n- Linux\u002FmacOS: \n```bash\nexport OPENAI_API_KEY=\u003Cyour-openai-api-key>\n```\n\n- Windows Command Prompt: \n```cmd \nset OPENAI_API_KEY=\u003Cyour-openai-api-key>\n```\n\n-  Windows PowerShell:\n```powershell\n$env:OPENAI_API_KEY=\"\u003Cyour-openai-api-key>\" # \" is required \n```\n\nOnce set, you can access the key in your Python code with:\n```python\nimport os\nOPENAI_API_KEY = os.getenv(\"OPENAI_API_KEY\")\n```\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>Option 2: Use .env File\u003C\u002Fsummary> \n\n- Create a .env file in your project root and add the following:\n```bash\nOPENAI_API_KEY=\u003Cyour-openai-api-key>\n```\n\nThen load it in Python:\n```python\nfrom dotenv import load_dotenv \nimport os \n\nload_dotenv() # Loads environment variables from .env file\nOPENAI_API_KEY = os.getenv(\"OPENAI_API_KEY\")\n```\n\u003C\u002Fdetails>\n\u003C!-- > 🔐 Tip: Don't forget to add `.env` to your `.gitignore` to avoid committing secrets. -->\n\n### Configure and Use the LLM\nOnce the API key is set, initialise the LLM with:\n\n```python\nfrom evoagentx.models import OpenAILLMConfig, OpenAILLM\n\n# Load the API key from environment\nOPENAI_API_KEY = os.getenv(\"OPENAI_API_KEY\")\n\n# Define LLM configuration\nopenai_config = OpenAILLMConfig(\n    model=\"gpt-4o-mini\",       # Specify the model name\n    openai_key=OPENAI_API_KEY, # Pass the key directly\n    stream=True,               # Enable streaming response\n    output_response=True       # Print response to stdout\n)\n\n# Initialize the language model\nllm = OpenAILLM(config=openai_config)\n\n# Generate a response from the LLM\nresponse = llm.generate(prompt=\"What is Agentic Workflow?\")\n```\n> 📖 More details on supported models and config options: [LLM module guide](.\u002Fdocs\u002Fmodules\u002Fllm.md).\n\n\n## Automatic WorkFlow Generation \nOnce your API key and language model are configured, you can automatically generate and execute multi-agent workflows in EvoAgentX.\n\n🧩 Core Steps:\n1. Define a natural language goal\n2. Generate the workflow with `WorkFlowGenerator`\n3. Instantiate agents using `AgentManager`\n4. Execute the workflow via `WorkFlow`\n\n💡 Minimal Example:\n```python\nfrom evoagentx.workflow import WorkFlowGenerator, WorkFlowGraph, WorkFlow\nfrom evoagentx.agents import AgentManager\n\ngoal = \"Generate html code for the Tetris game\"\nworkflow_graph = WorkFlowGenerator(llm=llm).generate_workflow(goal)\n\nagent_manager = AgentManager()\nagent_manager.add_agents_from_workflow(workflow_graph, llm_config=openai_config)\n\nworkflow = WorkFlow(graph=workflow_graph, agent_manager=agent_manager, llm=llm)\noutput = workflow.execute()\nprint(output)\n```\n\nYou can also:\n- 📊 Visualise the workflow: `workflow_graph.display()`\n- 💾 Save\u002Fload workflows: `save_module()` \u002F `from_file()`\n\n> 📂 For a complete working example, check out the [`workflow_demo.py`](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX\u002Fblob\u002Fmain\u002Fexamples\u002Fworkflow_demo.py)\n\n\n## 🧰 EvoAgentX Built-in Tools Summary\nEvoAgentX ships with a comprehensive suite of **built-in tools**, enabling agents to interact with code environments, search engines, databases, filesystems, images, and browsers. These modular toolkits form the backbone of multi-agent workflows and are easy to extend, customize, and test.\n\nCategories include:\n- 🧮 Code Interpreters (Python, Docker)\n- 🔍 Search & HTTP Requests (Google, Wikipedia, arXiv, RSS)\n- 🗂️ Filesystem Utilities (read\u002Fwrite, shell commands)\n- 🧠 Databases (MongoDB, PostgreSQL, FAISS)\n- 🖼️ Image Tools (analysis, generation)\n- 🌐 Browser Automation (low-level & LLM-driven)\n\nWe actively welcome contributions from the community!  \nFeel free to propose or submit new tools via [pull requests](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX\u002Fpulls) or [discussions](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX\u002Fdiscussions).\n\n\n\u003Cdetails>\n\u003Csummary>Click to expand full table 🔽\u003C\u002Fsummary>\n\n\u003Cbr>\n  \n| Toolkit Name | Description | Code File Path | Test File Path |\n|--------------|-------------|----------------|----------------|\n| **🧰 Code Interpreters** |  |  |  |\n| PythonInterpreterToolkit | Safely execute Python code snippets or local .py scripts with sandboxed imports and controlled filesystem access. | [link](evoagentx\u002Ftools\u002Finterpreter_python.py) | [link](examples\u002Ftools\u002Ftools_interpreter.py) |\n| DockerInterpreterToolkit | Run code (e.g., Python) inside an isolated Docker container—useful for untrusted code, special deps, or strict isolation. | [link](evoagentx\u002Ftools\u002Finterpreter_docker.py) | [link](examples\u002Ftools\u002Ftools_interpreter.py) |\n| **🧰 Search & Request Tools** |  |  |  |\n| WikipediaSearchToolkit | Search Wikipedia and retrieve results with title, summary, full content, and URL. | [link](evoagentx\u002Ftools\u002Fsearch_wiki.py) | [link](examples\u002Ftools\u002Ftools_search.py) |\n| GoogleSearchToolkit | Google Custom Search (official API). Requires GOOGLE_API_KEY and GOOGLE_SEARCH_ENGINE_ID. | [link](evoagentx\u002Ftools\u002Fsearch_google.py) | [link](examples\u002Ftools\u002Ftools_search.py) |\n| GoogleFreeSearchToolkit | Google-style search without API credentials (lightweight alternative). | [link](evoagentx\u002Ftools\u002Fsearch_google_f.py) | [link](examples\u002Ftools\u002Ftools_search.py) |\n| DDGSSearchToolkit | Search using DDGS with multiple backends and privacy-focused results | [link](evoagentx\u002Ftools\u002Fsearch_ddgs.py) | [link](examples\u002Ftools\u002Ftools_search.py) |\n| SerpAPIToolkit | Multi-engine search via SerpAPI (Google\u002FBing\u002FBaidu\u002FYahoo\u002FDDG) with optional content scraping. Requires SERPAPI_KEY. | [link](evoagentx\u002Ftools\u002Fsearch_serpapi.py) | [link](examples\u002Ftools\u002Ftools_search.py) |\n| SerperAPIToolkit | Google search via SerperAPI with content extraction. Requires SERPERAPI_KEY. | [link](evoagentx\u002Ftools\u002Fsearch_serperapi.py) | [link](examples\u002Ftools\u002Ftools_search.py) |\n| RequestToolkit | General HTTP client (GET\u002FPOST\u002FPUT\u002FDELETE) with params, form, JSON, headers, raw\u002Fprocessed response, and optional save to file. | [link](evoagentx\u002Ftools\u002Frequest.py) | [link](examples\u002Ftools\u002Ftools_search.py) |\n| ArxivToolkit | Search arXiv for research papers (title, authors, abstract, links\u002Fcategories). | [link](evoagentx\u002Ftools\u002Frequest_arxiv.py) | [link](examples\u002Ftools\u002Ftools_search.py) |\n| RSSToolkit | Fetch RSS feeds (with optional webpage content extraction) and validate feeds. | [link](evoagentx\u002Ftools\u002Frss_feed.py) | [link](examples\u002Ftools\u002Ftools_search.py) |\n| GoogleMapsToolkit | Geoinformation retrieval and path planning via Google API service. | [link](evoagentx\u002Ftools\u002Fgoogle_maps_tool.py) | [link](examples\u002Ftools\u002Fgoogle_maps_example.py) |\n| **🧰 FileSystem Tools** |  |  |  |\n| StorageToolkit | File I\u002FO utilities: save\u002Fread\u002Fappend\u002Fdelete, check existence, list files, list supported formats (pluggable storage backends). | [link](evoagentx\u002Ftools\u002Fstorage_file.py) | [link](examples\u002Ftools\u002Ftools_files.py) |\n| CMDToolkit | Execute shell\u002FCLI commands with working directory and timeout control; returns stdout\u002Fstderr\u002Freturn code. | [link](evoagentx\u002Ftools\u002Fcmd_toolkit.py) | [link](examples\u002Ftools\u002Ftools_files.py) |\n| FileToolkit | File operations toolkit for managing files and directories | [link](evoagentx\u002Ftools\u002Ffile_tool.py) | [link](examples\u002Ftools\u002Ftools_files.py) |\n| **🧰 Database Tools** |  |  |  |\n| MongoDBToolkit | MongoDB operations—execute queries\u002Faggregations, find with filter\u002Fprojection\u002Fsort, update, delete, info. | [link](evoagentx\u002Ftools\u002Fdatabase_mongodb.py) | [link](examples\u002Ftools\u002Ftools_database.py) |\n| PostgreSQLToolkit | PostgreSQL operations—generic SQL execution, targeted SELECT (find), UPDATE, CREATE, DELETE, INFO. | [link](evoagentx\u002Ftools\u002Fdatabase_postgresql.py) | [link](examples\u002Ftools\u002Ftools_database.py) |\n| FaissToolkit | Vector database (FAISS) for semantic search—insert documents (auto chunk+embed), query by similarity, delete by id\u002Fmetadata, stats. | [link](evoagentx\u002Ftools\u002Fdatabase_faiss.py) | [link](examples\u002Ftools\u002Ftools_database.py) |\n| **🧰 Image Handling Tools** |  |  |  |\n| ImageAnalysisToolkit | Vision analysis (OpenRouter GPT-4o family): describe images, extract objects\u002FUI info, answer questions about an image. | [link](evoagentx\u002Ftools\u002FOpenAI_Image_Generation.py) | [link](examples\u002Ftools\u002Ftools_images.py) |\n| OpenAIImageGenerationToolkit | Text-to-image via OpenAI (DALL·E family) with size\u002Fquality\u002Fstyle controls. | [link](evoagentx\u002Ftools\u002FOpenAI_Image_Generation.py) | [link](examples\u002Ftools\u002Ftools_images.py) |\n| FluxImageGenerationToolkit | Text-to-image via Flux Kontext Max (BFL) with aspect ratio, seed, format, prompt upsampling, and safety tolerance. | [link](evoagentx\u002Ftools\u002Fflux_image_generation.py) | [link](examples\u002Ftools\u002Ftools_images.py) |\n| **🧰 Browser Tools** |  |  |  |\n| BrowserToolkit | Fine-grained browser automation: initialize, navigate, type, click, resnapshot page, read console logs, and close. | [link](evoagentx\u002Ftools\u002Fbrowser_tool.py) | [link](examples\u002Ftools\u002Ftools_browser.py) |\n| BrowserUseToolkit | High-level, natural-language browser automation (navigate, fill forms, click, search, etc.) driven by an LLM. | [link](evoagentx\u002Ftools\u002Fbrowser_use.py) | [link](examples\u002Ftools\u002Ftools_browser.py) |\n\n\u003C\u002Fdetails>\n\n**EvoAgentX also supports MCP tools.**  \nCheck out our [tutorial](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX\u002Fblob\u002Fmain\u002Fdocs\u002Ftutorial\u002Fmcp.md) to learn how to configure your preferred MCP tools with EvoAgentX.\n\n## Tool-Enabled Workflows Generation:\n\nIn more advanced scenarios, your workflow agents may need to use external tools. EvoAgentX allows Automatic tool integration: Provide a list of toolkits to WorkFlowGenerator. The generator will consider these and include them in the agents if appropriate.\n\nFor instance, to enable an Arxiv toolkit:\n```python\nfrom evoagentx.tools import ArxivToolkit\n\n# Initialize a command-line toolkit for file operations\narxiv_toolkit = ArxivToolkit()\n\n# Generate a workflow with the toolkit available to agents\nwf_generator = WorkFlowGenerator(llm=llm, tools=[arxiv_toolkit])\nworkflow_graph = wf_generator.generate_workflow(goal=\"Find and summarize the latest research on AI in the field of finance on arXiv\")\n\n# Instantiate agents with access to the toolkit\nagent_manager = AgentManager(tools=[arxiv_toolkit])\nagent_manager.add_agents_from_workflow(workflow_graph, llm_config=openai_config)\n\nworkflow = WorkFlow(graph=workflow_graph, agent_manager=agent_manager, llm=llm)\noutput = workflow.execute()\nprint(output)\n```\n\nIn this setup, the workflow generator may assign the `ArxivToolkit` to relevant agents, enabling them to execute shell commands as part of the workflow (e.g. creating directories and files)\n\n## Human-in-the-Loop (HITL) Support:\n\nIn advanced scenarios, EvoAgentX supports integrating human-in-the-loop interactions within your agent workflows. This means you can pause an agent’s execution for manual approval or inject user-provided input at key steps, ensuring critical decisions are vetted by a human when needed.\n\nAll human interactions are managed through a central `HITLManager` instance. The HITL module includes specialized agents like `HITLInterceptorAgent` for approval gating and `HITLUserInputCollectorAgent` for collecting user data.\n\nFor instance, to require human approval before an email-sending agent executes its action:\n```python\nfrom evoagentx.hitl import HITLManager, HITLInterceptorAgent, HITLInteractionType, HITLMode\n\nhitl_manager = HITLManager()\nhitl_manager.activate()  # Enable HITL (disabled by default)\n\n# Interceptor agent to approve\u002Freject the DummyEmailSendAction of DataSendingAgent\ninterceptor = HITLInterceptorAgent(\n    target_agent_name=\"DataSendingAgent\",\n    target_action_name=\"DummyEmailSendAction\",\n    interaction_type=HITLInteractionType.APPROVE_REJECT,\n    mode=HITLMode.PRE_EXECUTION    # ask before action runs\n)\n# Map the interceptor’s output field back to the workflow’s input field for continuity\nhitl_manager.hitl_input_output_mapping = {\"human_verified_data\": \"extracted_data\"}\n\n# Add the interceptor to the AgentManager and include HITL in the workflow execution\nagent_manager.add_agent(interceptor)\nworkflow = WorkFlow(graph=workflow_graph, agent_manager=agent_manager, llm=llm, hitl_manager=hitl_manager)\n```\nWhen this interceptor triggers, the workflow will pause and prompt in the console for `[a]pprove` or `[r]eject` before continuing. If approved, the flow proceeds using the human-verified data; if rejected, the action is skipped or handled accordingly.\n\n> 📂 For a complete working example, check out the [`tutorial\n\u002Fhitl.md`](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX\u002Fblob\u002F615b06d29264f47e58a6780bd24f0e73cbf7deee\u002Fdocs\u002Ftutorial\u002Fhitl.md)\n\n## Demo Video\n\n\n[![Watch on YouTube](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F-Watch%20on%20YouTube-red?logo=youtube&labelColor=grey)](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=8ALcspHOe0o)\n[![Watch on Bilibili](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F-Watch%20on%20Bilibili-00A1D6?logo=bilibili&labelColor=white)](https:\u002F\u002Fwww.bilibili.com\u002Fvideo\u002FBV1AjahzRECi\u002F?vd_source=02f8f3a7c8865b3af6378d9680393f5a)\n\n\u003Cdiv align=\"center\">\n  \u003Cvideo src=\"https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002F65af8cce-43ad-4e81-ab8d-fc085a7fdc05.mp4\" autoplay loop muted playsinline width=\"600\">\n    Your browser does not support the video tag.\n  \u003C\u002Fvideo>\n\u003C\u002Fdiv>\n\nIn this demo, we showcase the workflow generation and execution capabilities of EvoAgentX through two examples:\n\n- **Application 1: Financial Information Agentic Workflow**\n- \n  In this example, we use a workflow generated by EvoAgentX to collect public information about a company based on a given index.  \n  The collected data includes the overall market index, the company’s current stock price, institutional buy\u002Fsell activity, and more.  \n  Finally, the workflow generates an **HTML report** summarizing the information and providing a buy\u002Fsell\u002Fhold recommendation. This workflow is only an alpha version.\n  If you're interested in turning it into a **truly practical investment assistant**, you can consider integrating more financial indicators and analytical tools—and let these tools join your\n  workflow through agents! Check [here](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX\u002Fblob\u002Fmain\u002Fexamples\u002Fworkflow\u002Finvest\u002Fstock_analysis.py) to try this workflow.\n  \n- **Application 2: ArXiv Research Summarizer Workflow**\n\n  This workflow, generated by EvoAgentX and powered by the ArXiv MCP tool, can retrieve and summarize relevant papers from arXiv based on your provided keywords and selected time range.  \n  If you're interested, you can even **extend this workflow beyond arXiv**, integrating it with other academic search platforms like **Google Scholar**, and turn it into a fully functional research assistant application! Check [here](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX\u002Fblob\u002Fmain\u002Fexamples\u002Fworkflow\u002Farxiv_workflow.py) to play with this workflow. \n\n### ✨ Final Results\n\n\u003Ctable>\n  \u003Ctr>\n    \u003Ctd align=\"center\">\n      \u003Cimg src=\".\u002Fassets\u002Fdemo_result_1.png\" width=\"400\">\u003Cbr>\n      \u003Cstrong>Application 1:\u003C\u002Fstrong>\u003Cbr>Stock Recommendation\n    \u003C\u002Ftd>\n    \u003Ctd align=\"center\">\n      \u003Cimg src=\".\u002Fassets\u002Fdemo_result_2.png\" width=\"400\">\u003Cbr>\n      \u003Cstrong>Application 2:\u003C\u002Fstrong>\u003Cbr>Arxiv Daily Paper Recommendation\n    \u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\n## Evolution Algorithms \n\nWe have integrated some effective agent\u002Fworkflow evolution algorithms into EvoAgentX:\n\n| **Algorithm** | **Description** | **Link** |\n|---------------|-----------------|----------|\n| **TextGrad**  | Gradient-based optimization for LLM prompts and reasoning chains, enabling differentiable planning. | [📄 Nature (2025)](https:\u002F\u002Fwww.nature.com\u002Farticles\u002Fs41586-025-08661-4) |\n| **MIPRO**     | Model-agnostic Iterative Prompt Optimization using black-box evaluations and adaptive reranking. | [📄 arXiv:2406.11695](https:\u002F\u002Farxiv.org\u002Fabs\u002F2406.11695) |\n| **AFlow**     | Reinforcement learning-inspired agent workflow evolution using Monte Carlo Tree Search. | [📄 arXiv:2410.10762](https:\u002F\u002Farxiv.org\u002Fabs\u002F2410.10762) |\n| **EvoPrompt**     | EvoPrompt dynamically refines prompts via feedback-driven evolution to enhance agent performance and adaptability. | [📄 arXiv:2309.08532](https:\u002F\u002Farxiv.org\u002Fabs\u002F2309.08532) |\n\n\n**Please suggest the latest self-evolving algorithm by submitting an issue or a Pull Request.**\n\nTo evaluate the performance, we use them to optimize the same agent system on three different tasks: multi-hop QA (HotPotQA), code generation (MBPP) and reasoning (MATH). We randomly sample 50 examples for validation and other 100 examples for testing. \n\n> Tip: We have integrated these benchmark and evaluation code in EvoAgentX. Please refer to the [benchmark and evaluation tutorial](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX\u002Fblob\u002Fmain\u002Fdocs\u002Ftutorial\u002Fbenchmark_and_evaluation.md) for more details.\n\n### 📊 Results \n\n| Method   | HotPotQA\u003Cbr>(F1%) | MBPP\u003Cbr>(Pass@1 %) | MATH\u003Cbr>(Solve Rate %) |\n|----------|--------------------|---------------------|--------------------------|\n| Original | 63.58              | 69.00               | 66.00                    |\n| TextGrad | 71.02              | 71.00               | 76.00                    |\n| AFlow    | 65.09              | 79.00               | 71.00                    |\n| MIPRO    | 69.16              | 68.00               | 72.30       \n\nPlease refer to the `examples\u002Foptimization` folder for more details. \n\n## Applications \n\nWe use our framework to optimize existing multi-agent systems on the [GAIA](https:\u002F\u002Fhuggingface.co\u002Fspaces\u002Fgaia-benchmark\u002Fleaderboard) benchmark. We select [Open Deep Research](https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Fsmolagents\u002Ftree\u002Fmain\u002Fexamples\u002Fopen_deep_research) and [OWL](https:\u002F\u002Fgithub.com\u002Fcamel-ai\u002Fowl), two representative multi-agent framework from the GAIA leaderboard that is open-source and runnable. \n\nWe apply EvoAgentX to optimize their prompts. The performance of the optimized agents on the GAIA benchmark validation set is shown in the figure below.\n\n\u003Ctable>\n  \u003Ctr>\n    \u003Ctd align=\"center\" width=\"50%\">\n      \u003Cimg src=\".\u002Fassets\u002Fopen_deep_research_optimization_report.png\" alt=\"Open Deep Research Optimization\" width=\"100%\">\u003Cbr>\n      \u003Cstrong>Open Deep Research\u003C\u002Fstrong>\n    \u003C\u002Ftd>\n    \u003Ctd align=\"center\" width=\"50%\">\n      \u003Cimg src=\".\u002Fassets\u002Fowl_optimization_result.png\" alt=\"OWL Optimization\" width=\"100%\">\u003Cbr>\n      \u003Cstrong>OWL Agent\u003C\u002Fstrong>\n    \u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\n> Full Optimization Reports: [Open Deep Research](https:\u002F\u002Fgithub.com\u002Feax6\u002Fsmolagents) and [OWL](https:\u002F\u002Fgithub.com\u002FTedSIWEILIU\u002Fowl).  \n\n## Tutorial and Use Cases\n\n> 💡 **New to EvoAgentX?** Start with the [Quickstart Guide](.\u002Fdocs\u002Fquickstart.md) for a step-by-step introduction.\n\n\nExplore how to effectively use EvoAgentX with the following resources:\n\n| Cookbook | Colab Notebook | Description |\n|:---|:---|:---|\n| **[Build Your First Agent](.\u002Fdocs\u002Ftutorial\u002Ffirst_agent.md)** | **[Build Your First Agent](.\u002Fdocs\u002FColabNotebook\u002Ftutorial_notebooks\u002Ffirst_agent.ipynb)** | Quickly create and manage agents with multi-action capabilities. |\n| **[Build Your First Workflow](.\u002Fdocs\u002Ftutorial\u002Ffirst_workflow.md)** | **[Build Your First Workflow](.\u002Fdocs\u002FColabNotebook\u002Ftutorial_notebooks\u002Ffirst_workflow.ipynb)** | Learn to build collaborative workflows with multiple agents. |\n| **[Working with Tools](.\u002Fdocs\u002Ftutorial\u002Ftools.md)** | **[Working with Tools](.\u002Fdocs\u002FColabNotebook\u002Ftutorial_notebooks\u002Ftools.ipynb)** | Master EvoAgentX's powerful tool ecosystem for agent interactions |\n| **[Automatic Workflow Generation](.\u002Fdocs\u002Fquickstart.md#automatic-workflow-generation-and-execution)** | **[Automatic Workflow Generation](.\u002Fdocs\u002FColabNotebook\u002Ftutorial_notebooks\u002Fquickstart.ipynb)** | Automatically generate workflows from natural language goals. |\n| **[Benchmark and Evaluation Tutorial](.\u002Fdocs\u002Ftutorial\u002Fbenchmark_and_evaluation.md)** | **[Benchmark and Evaluation Tutorial](.\u002Fdocs\u002FColabNotebook\u002Ftutorial_notebooks\u002Fbenchmark_and_evaluation.ipynb)** | Evaluate agent performance using benchmark datasets. |\n| **[TextGrad Optimizer Tutorial](.\u002Fdocs\u002Ftutorial\u002Ftextgrad_optimizer.md)** | **[TextGrad Optimizer Tutorial](.\u002Fdocs\u002FColabNotebook\u002Ftutorial_notebooks\u002Ftextgrad_optimizer.ipynb)** | Automatically optimise the prompts within multi-agent workflow with TextGrad. |\n| **[AFlow Optimizer Tutorial](.\u002Fdocs\u002Ftutorial\u002Faflow_optimizer.md)** | **[AFlow Optimizer Tutorial](.\u002Fdocs\u002FColabNotebook\u002Ftutorial_notebooks\u002Faflow_optimizer.ipynb)** | Automatically optimise both the prompts and structure of multi-agent workflow with AFlow. |\n| **[Human-In-The-Loop support](.\u002Fdocs\u002Ftutorial\u002Fhitl.md)** | | Enable HITL functionalities in your WorkFlow.\n\u003C!-- | **[SEW Optimizer Tutorial](.\u002Fdocs\u002Ftutorial\u002Fsew_optimizer.md)** | Create SEW (Self-Evolving Workflows) to enhance agent systems. | -->\n\n🛠️ Follow the tutorials to build and optimize your EvoAgentX workflows.\n\n🚀 We're actively working on expanding our library of use cases and optimization strategies. **More coming soon — stay tuned!**\n\n\n## 🗣️ EvoAgentX TALK\n\nEvoAgentX regularly invites leading researchers to give guest lectures on cutting-edge AI topics.  \nBelow is a running log of scheduled and completed talks:\n\n| Speaker | Topic | Date | Meeting Video |\n|---|---|---|---|\n| [Wang Hongru](https:\u002F\u002Fhrwise-nlp.github.io\u002F) | [EvoAgentX Talk] Second Half of Languages Agent | 2026-05-10 | [YouTube](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=p8HU_lFtAM4)\n| [Zhou Huichi](https:\u002F\u002Fscholar.google.com\u002Fcitations?user=1IJyxpUAAAAJ&hl=en) | [EvoAgentX Talk] Momentum and Continual Learning Agent Systems | 2026-04-26 | [YouTube](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=DHKBq4NYAxA&t=53s)\n| [Xi Zhang](https:\u002F\u002Fx-izhang.github.io\u002F) | [EvoAgentX Talk] EvoScientist | 2026-04-05 |  [YouTube](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=EVrxBC99_1k&t=2745s)  |\n| [Guanting Dong](https:\u002F\u002Fdongguanting.github.io\u002F) | [EvoAgentX Talk] Agentic Reinforced Policy Optimization | 2025-10-30 |  [YouTube](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=UB-FL5JFXJY)  |\n| [Guibin Zhang](https:\u002F\u002Fwww.guibinz.top\u002F) | [EvoAgentX Talk] How Agentic RL Transforms LLMs into Automated Agents | 2025-09-28 | [YouTube](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=xzqZsZWfabw&t=3s) |\n| [EvoAgentX Team](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX) | [EvoAgentX Demo] Self-Evolving Agents in Action: Stock Analysis Assistant & arXiv Paper Recommender | 2025-09-04 | [YouTube](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=8ALcspHOe0o&t=33s) |\n| [EvoAgentX Team](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX) | [EvoAgentX Meeting] EvoAgentX Open-Source Meeting:2025-08-17 | 2025-08-17 | [YouTube](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=Xy4sObew-a0&t=9s) |\n| [Hengzhe Zhang](https:\u002F\u002Fhengzhe-zhang.github.io\u002F) | [EvoAgentX Talk] Genetic Programming: From Evolutionary Algorithms to the LLM Era | 2025-08-10 | [YouTube](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=naja_kDYc_Y) |\n| [EvoAgentX Team](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX) | [EvoAgentX Meeting] EvoAgentX Open-Source Meeting: 2025-07-27 | 2025-07-27 | [YouTube](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=dgJFCrbllWI) |\n| [EvoAgentX Team](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX) | [EvoAgentX Meeting] EvoAgentX Open-Source Meeting: 2025-07-20 | 2025-07-20 | [YouTube](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=mnJYmdgo8HA&t=68s) |\n| [EvoAgentX Team](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX) | [EvoAgentX Meeting] EvoAgentX Open-Source Meeting: 2025-07-06 | 2025-07-06 | [YouTube](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=AIAl6rB4fC0) |\n| [EvoAgentX Team](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX) | [EvoAgentX Meeting] EvoAgentX Open-Source Meeting: 2025-06-22 | 2025-06-22 | [YouTube](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=3y0Ppq3BN1A&t=1401s) |\n| [EvoAgentX Team](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX) | [EvoAgentX Meeting] EvoAgentX First Community Call: Exploring the Future of Self-Evolving AI Agents! | 2025-06-10 | [YouTube](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=WzWo9aYPHMs&t=414s) |\n| [EvoAgentX Team](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX) | [EvoAgentX Demo] EvoAgentX Workflow Generation Demo | 2025-05-14 | [YouTube](https:\u002F\u002Fwww.youtube.com\u002Fwatch?v=Wu0ZydYDqgg) |\n\n\n## 🎯 Roadmap\n- [ ] **Modularize Evolution Algorithms**: Abstract optimization algorithms into plug-and-play modules that can be easily integrated into custom workflows. \n- [ ] **Develop Task Templates and Agent Modules**: Build reusable templates for typical tasks and standardized agent components to streamline application development.\n- [ ] **Integrate Self-Evolving Agent Algorithms**: Incorporate more recent and advanced agent self-evolution across multiple dimensions, including prompt tuning, workflow structures, and memory modules. \n- [ ] **Enable Visual Workflow Editing Interface**: Provide a visual interface for workflow structure display and editing to improve usability and debugging. \n\n\n## 🙋 Support\n\n### Join the Community\n\n📢 Stay connected and be part of the **EvoAgentX** journey!  \n🚩 Join our community to get the latest updates, share your ideas, and collaborate with AI enthusiasts worldwide.\n\n- [Discord](https:\u002F\u002Fdiscord.gg\u002FXWBZUJFwKe) — Chat, discuss, and collaborate in real-time.\n- [X (formerly Twitter)](https:\u002F\u002Fx.com\u002FEvoAgentX) — Follow us for news, updates, and insights.\n- [WeChat](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX\u002Fblob\u002Fmain\u002Fassets\u002Fwechat_info.md) — Connect with our Chinese community.\n\n### Add the meeting to your calendar\n\n📅 Click the link below to add the EvoAgentX Weekly Meeting (Sundays, 16:30–17:30 GMT+8) to your calendar:\n\n👉 [Add to your Google Calendar](https:\u002F\u002Fcalendar.google.com\u002Fcalendar\u002Fu\u002F0\u002Fr\u002Feventedit?text=EvoAgentX+周会（腾讯会议）&dates=20250629T083000Z\u002F20250629T093000Z&details=会议链接：https:\u002F\u002Fmeeting.tencent.com\u002Fdm\u002F5UuNxo7Detz0&location=Online&recur=RRULE:FREQ=WEEKLY;BYDAY=SU;UNTIL=20270523T093000Z&ctz=Asia\u002FShanghai)\n\n👉 [Add to your Tencent Meeting](https:\u002F\u002Fmeeting.tencent.com\u002Fdm\u002F5UuNxo7Detz0)\n\n👉 [Download the EvoAgentX_Weekly_Meeting.ics file](.\u002FEvoAgentX_Weekly_Meeting.ics)\n\n### Contact Information\n\nIf you have any questions or feedback about this project, please feel free to contact us. We highly appreciate your suggestions!\n\n- **Email:** evoagentx.ai@gmail.com\n\nWe will respond to all questions within 2-3 business days.\n\n### Community Call\n- [Bilibili](https:\u002F\u002Fspace.bilibili.com\u002F3493105294641286\u002Ffavlist?fid=3584589186&ftype=create&spm_id_from=333.788.0.0)\n- [Youtube](https:\u002F\u002Fstudio.youtube.com\u002Fplaylist\u002FPL_kuPS05qA1hyU6cLX--bJ93Km2-md8AA\u002Fedit)\n## 🙌 Contributing to EvoAgentX\nThanks go to these awesome contributors\n\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX\u002Fgraphs\u002Fcontributors\">\n  \u003Cimg src=\"https:\u002F\u002Fcontrib.rocks\u002Fimage?repo=EvoAgentX\u002FEvoAgentX\" \u002F>\n\u003C\u002Fa>\n\nWe appreciate your interest in contributing to our open-source initiative. We provide a document of [contributing guidelines](https:\u002F\u002Fgithub.com\u002FEvoAgentX\u002FEvoAgentX\u002Fblob\u002Fmain\u002FCONTRIBUTING.md) which outlines the steps for contributing to EvoAgentX. Please refer to this guide to ensure smooth collaboration and successful contributions. 🤝🚀\n\n[![Star History Chart](https:\u002F\u002Fapi.star-history.com\u002Fsvg?repos=EvoAgentX\u002FEvoAgentX&type=Date)](https:\u002F\u002Fwww.star-history.com\u002F#EvoAgentX\u002FEvoAgentX&Date)\n\n## 📖 Citation\n\nPlease consider citing our work if you find EvoAgentX helpful:\n\n📄 [EvoAgentX](https:\u002F\u002Farxiv.org\u002Fabs\u002F2507.03616)\n📄 [Survey Paper](https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.07407)\n\n```bibtex\n@article{wang2025evoagentx,\n  title={EvoAgentX: An Automated Framework for Evolving Agentic Workflows},\n  author={Wang, Yingxu and Liu, Siwei and Fang, Jinyuan and Meng, Zaiqiao},\n  journal={arXiv preprint arXiv:2507.03616},\n  year={2025}\n}\n@article{fang202survey,\n      title={A Comprehensive Survey of Self-Evolving AI Agents: A New Paradigm Bridging Foundation Models and Lifelong Agentic Systems}, \n      author={Jinyuan Fang and Yanwen Peng and Xi Zhang and Yingxu Wang and Xinhao Yi and Guibin Zhang and Yi Xu and Bin Wu and Siwei Liu and Zihao Li and Zhaochun Ren and Nikos Aletras and Xi Wang and Han Zhou and Zaiqiao Meng},\n      year={2025},\n      journal={arXiv preprint arXiv:2508.07407},\n      url={https:\u002F\u002Farxiv.org\u002Fabs\u002F2508.07407}, \n}\n```\n\n## 📚 Acknowledgements \nThis project builds upon several outstanding open-source projects: [AFlow](https:\u002F\u002Fgithub.com\u002FFoundationAgents\u002FMetaGPT\u002Ftree\u002Fmain\u002Fmetagpt\u002Fext\u002Faflow), [TextGrad](https:\u002F\u002Fgithub.com\u002Fzou-group\u002Ftextgrad), [DSPy](https:\u002F\u002Fgithub.com\u002Fstanfordnlp\u002Fdspy), [EvoPrompt](https:\u002F\u002Fgithub.com\u002Fbeeevita\u002FEvoPrompt), [LiveCodeBench](https:\u002F\u002Fgithub.com\u002FLiveCodeBench\u002FLiveCodeBench)and more. We would like to thank the developers and maintainers of these frameworks for their valuable contributions to the open-source community.\n\n## 📄 License\n\nSource code in this repository is made available under the [MIT License](.\u002FLICENSE).\n","EvoAgentX 是一个用于构建、评估和进化基于大语言模型的代理或代理工作流的开源框架。它支持自动化、模块化及目标驱动的方式，使开发者和研究人员能够超越静态提示链或手动工作流编排的限制。项目的核心功能包括从单一提示构建结构化的多代理工作流、内置评估器自动评分代理行为以及通过自我进化算法优化工作流。此外，EvoAgentX 提供了即插即用的兼容性，便于集成 OpenAI 和 Qwen 等多种模型。该项目适用于需要动态调整和持续优化的多代理系统场景，如复杂的自然语言处理任务、信息检索增强生成（RAG）等。",2,"2026-06-11 03:41:32","high_star"]