EvoAgentX

What it is

EvoAgentX is an open-source framework for building, evaluating, and evolving LLM-based agents or agentic workflows in an automated, modular, and goal-driven manner. It aims to move beyond static prompt chaining and manual workflow orchestration by enabling a self-evolving ecosystem of AI agents.

Key features

• Agent Workflow Autoconstruction: From a single prompt, EvoAgentX builds structured, multi-agent workflows tailored to the task.
• Built-in Evaluation: Integrates automatic evaluators to score agent behavior using task-specific criteria.
• Self-Evolution Engine: Agents learn and improve workflows through self-evolving algorithms.
• Plug-and-Play Compatibility: Integrates original OpenAI and qwen models, and supports other models via additional adapters.
• Built-in Tools: Includes a suite of tools for interacting with code, search, databases, filesystems, images, and browsers.
• Memory Module: Supports ephemeral and persistent memory systems.
• Human-in-the-Loop (HITL): Supports interactive workflows where humans review and guide behavior.

How it works

The framework orchestrates multi-agent workflows by automatically generating workflows from natural language goals, instantiating agents, and executing the workflow. It provides a minimal example showing generation of a workflow from a goal, creation of an AgentManager, and execution of a WorkFlow object. It also includes a workflow graph visualization and save/load capabilities.

Getting started

• Installation:

pip install evoagentx

or install from source:

pip install git+https://github.com/EvoAgentX/EvoAgentX.git

For local development, users can follow the provided example to clone the repo, create a conda environment, and install requirements or install in development mode. The README includes an example:

git clone https://github.com/EvoAgentX/EvoAgentX.git
cd EvoAgentX
# Create a new conda environment
conda create -n evoagentx python=3.11

# Activate the environment
conda activate evoagentx

# Install the package
pip install -r requirements.txt
# OR install in development mode
pip install -e .

• LLM configuration:

from evoagentx.models import OpenAILLMConfig, OpenAILLM
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
openai_config = OpenAILLMConfig(
    model="gpt-4o-mini",
    openai_key=OPENAI_API_KEY,
    stream=True,
    output_response=True
)
llm = OpenAILLM(config=openai_config)
response = llm.generate(prompt="What is Agentic Workflow?")

• API key configuration options:

export OPENAI_API_KEY=<your-openai-api-key>

set OPENAI_API_KEY=<your-openai-api-key>

$env:OPENAI_API_KEY="<your-openai-api-key>"

• Also supports a .env file approach:

OPENAI_API_KEY=<your-openai-api-key>

from dotenv import load_dotenv 
import os 
load_dotenv()
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")

Recent releases

• v0.1.4 Release (2026-06-28): Improves workflow orchestration, parameter validation, and overall compatibility. Highlights include improved workflow execution and graph handling, with better compatibility across o
• v0.1.3 Release (2026-06-27): Focuses on agent/tool-calling upgrade and provider compatibility improvements. Highlights include refactored CustomizeAgent and CustomizeAction to support structu
• v0.1.2 Release (2026-06-24): Upgrades core LLM provider support and improves structured output handling. LLM Provider Updates include updated OpenAILLM and OpenRouterLLM to s
• v0.1.1 Release (2026-06-23): Enhancements in prompt templating and structured output parsing with JSON Schema support. Fixed module serialization and config restoration issues for ne
• v0.1.0 Initial Release (2025-09-06): First official version of the self-evolving agent framework.

Traction

Stars: 3206, Forks: 287, Open issues: 17

Behind the repo

The repository targets Python and is MIT-licensed per the repository header links; it includes a broad tool catalog and integration points for LLMs and local models.

Caveats

License: none listed in the provided FACTS. Created: 2025-04-15. Last push: 2026-07-07. Language: Python.