ProAgent
VerifiedLLM agent that builds and runs adaptive workflows from simple instructions.
What is ProAgent?
ProAgent is an open-source LLM agent that implements Agentic Process Automation. It accepts natural-language goals, generates executable workflows, and manages execution by calling specialized helper agents when decisions arise.
The system records every step so runs can be replayed, refined, or inspected later. It connects to real services through a self-hosted n8n instance or replays saved traces without external accounts.
Intended for developers and automation engineers who need reliable, inspectable workflows that adapt to changing conditions without constant manual redesign.
Capabilities
What you can build with ProAgent
Workflow generation from instructions
Turn a short description of a business process into a working n8n workflow that can be tested and adjusted.
Dynamic decision handling
Delegate runtime choices to specialized agents so the main workflow continues without human intervention at every branch.
Reproducible automation experiments
Load saved records to rerun or refine past executions in production or refine mode without repeating setup steps.
Install ProAgent
pip install -r requirements.txtpip install -r requirements.txt- 1Clone the repository and install dependencies with pip install -r requirements.txt using the supported OpenAI library version.
- 2Install and start a self-hosted n8n instance, then export decrypted credentials and workflows to the required JSON files.
- 3Set OPENAI_API_KEY and OPENAI_API_BASE environment variables if running in development mode.
- 4Edit config.py to choose development, refine, or production mode, then launch with python main.py.
- 5Review the generated record in the records folder to inspect, refine, or replay the automation run.
Works with
ProAgent: pros & cons
Pros
- +Produces human-readable records of every workflow construction and execution step.
- +Supports both live n8n connections and offline replay of published examples.
- +Allows iterative refinement of existing workflows through natural-language requests.
- +Demonstrates concrete end-to-end automation cases from the accompanying research paper.
Cons
- –Requires a working n8n installation and credential export, which can be error-prone.
- –Locked to an older OpenAI library version that may conflict with current installations.
- –Live webhook redirect service is external and noted as potentially unstable.
Frequently asked questions
Development mode uses live n8n; refine and production modes can load saved records without it.
User reviews
Verified reviews from the community shape this listing's rating.
Loading reviews…
Similar agents
Other automation options worth comparing.
