AgentFactory: A Self-Evolving Framework Through Executable Subagent Accumulation and Reuse

A Meta-Agent system that orchestrates problem-solving by decomposing complex tasks and delegating them to specialized subagents. The system features a self-evolving capability: subagents created during one task can be automatically reused and improved in subsequent tasks.

Setup

1. Install Dependencies

conda create -n agentfactory python=3.12
conda activate agentfactory
pip install openai requests playwright flask flask_cors

2. Configure API Key

Edit llm.py and set your API key and model configuration:

# For Claude
LLM_URL_CLAUDE = "https://your-api-endpoint/v1"
LLM_API_KEY_CLAUDE = "your-api-key"
LLM_MODEL_CLAUDE = "claude-opus-4-6"

# For MiniMax (local or remote)
LLM_URL_MINIMAX = "https://your-api-endpoint/v1"
LLM_API_KEY_MINIMAX = "your-api-key"
LLM_MODEL_MINIMAX = "MiniMax-M2.5"

Switch between models by changing model_choice at the top of llm.py.

3. Configure Tool API Keys

Edit tools.py and set your API keys for web search and page reading:

SERPER_API_KEY = "your-serper-api-key"   # Get from https://serper.dev
JINA_API_KEY = "your-jina-api-key"       # Get from https://jina.ai

4. Install Chrome

Browser-based tasks (e.g., web automation via Playwright) require Google Chrome. Make sure Chrome is installed on your machine before running these tasks.

5. Replace Placeholder Paths

In data/questions_round1.jsonl and data/questions_round2.jsonl, replace all occurrences of <Absolute_Path_to_AgentFactory> with the actual absolute path to this directory. For example:

<Absolute_Path_to_AgentFactory>/qq_music_taylor.mp3

should become:

/home/user/AgentFactory/qq_music_taylor.mp3

Running Tests

python run.py --question-file data/questions_round1.jsonl

Optional flags:

• --human-confirm — pause for human confirmation when finishing
• --no-save — do not save subagent skills after completion

Web Demo

Start the web server:

python3 web_interface/app.py --port 5050

Then open http://localhost:5050 in your browser. To have the agent solve a task described in an audio file, enter:

Help me complete a task. The detailed description of the task is in the audio file <your_absolute_path_to_audio>

For example, using the included sample audio:

Help me complete a task. The detailed description of the task is in the audio file /home/user/AgentFactory/tencent_doc_en.mp3

Replace the path with the actual absolute path to the audio file on your machine.

Visualizing Trajectories

Open visualize_trajectory.html in a browser, then select a trajectory JSON file to visualize the agent's execution process.

Recorded Demonstrations

The trajectory/ directory contains saved execution trajectories that correspond to the demonstrations referenced in the paper:

• trajectory/qq_music/ — Two runs of the QQ Music task.

  • first_time.json — First attempt; the agent creates subagents from scratch.
  • second_time.json — Second attempt; the agent reuses subagents created in the first run.

• trajectory/tencent_doc/ — Three runs of the Tencent Docs task.

  • first_time.json — First attempt with a text-based instruction; the agent creates subagents from scratch.
  • second_time.json — Second attempt with text; the agent reuses previously created subagents.
  • third_time.json — Third attempt with an audio-based instruction. After the QQ Music runs had already produced an audio_transcriber subagent, the agent directly reuses it here to transcribe the audio and solve the task.

• trajectory/readme/ — Three sequential runs demonstrating the self-evolution process, where the agent progressively builds and refines its subagent repertoire across tasks.