Skillget

agent-generator

mcp
Guvenlik Denetimi
Uyari
Health Uyari
  • License — License: NOASSERTION
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 5 GitHub stars
Code Gecti
  • Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Gecti
  • Permissions — No dangerous permissions requested
Purpose
This tool provides a CLI and web interface to transform plain-English prompts into complete, multi-agent AI workflow code. It generates native project files for frameworks like CrewAI, LangGraph, and IBM WatsonX Orchestrate, and includes a built-in FastAPI MCP server wrapper for deployment.

Security Assessment
Overall Risk: Medium. The application itself does not contain hardcoded secrets or request dangerous system permissions. However, a light code scan of 12 files may not capture the full risk profile of its dependencies. The tool processes sensitive data by design; it requires you to export API keys (such as OPENAI_API_KEY or WATSONX_API_KEY) into your environment variables to function. While standard practice for AI tools, this inherently exposes sensitive credentials to the application's runtime. Additionally, because it dynamically generates and executes Python code based on user prompts, there is an underlying risk of executing unintended or malicious instructions if the generated output is run without prior review.

Quality Assessment
The project is active and recently updated (last pushed 0 days ago). It has a proper description and claims an Apache 2.0 license, though the automated scanner flagged the license as "NOASSERTION" due to repository metadata. Community trust and visibility are currently very low, with only 5 GitHub stars. Consequently, the user base is small, and the project has not undergone widespread community auditing.

Verdict
Use with caution: the code appears clean and active, but its low community adoption, requirement for API keys, and dynamic code execution capabilities warrant a thorough review of generated outputs before running them.
SUMMARY

CLI and Flask‑based web application that transforms plain‑English prompts into production‑ready, multi‑agent AI workflows. It generates native YAML for IBM WatsonX Orchestrate, Python code for CrewAI, CrewAI Flow, LangGraph, or ReAct, and includes a built‑in FastAPI MCP server wrapper for seamless deployment to the MCP Gateway.

README.md

Agent-Generator Logo

agent-generator

Create AI agents from a single sentence.

PyPI
Python 3.10+
License
Try it
Powered by matrix-hub

What does it do?

You type what you want. You get a complete project.

agent-generator "Build a research team with a researcher and a writer" -f crewai

That's it. You now have working Python code with agents, tasks, and tools.

Try it online

No install needed. Open the demo and start building:

https://huggingface.co/spaces/ruslanmv/agent-generator

Install

pip install agent-generator

That gives you the CLI and all core features. Python 3.10+ required.

Optional extras:

pip install "agent-generator[openai]"     # OpenAI provider
pip install "agent-generator[all]"        # All providers + frameworks

Quick start

Step 1 -- Set credentials (pick one)

# Option A: WatsonX (default)
export WATSONX_API_KEY=your-key
export WATSONX_PROJECT_ID=your-project-id

# Option B: OpenAI
export OPENAI_API_KEY=sk-your-key
export AGENTGEN_PROVIDER=openai

Step 2 -- Generate

# CrewAI team
agent-generator "Research team that finds papers and writes summaries" -f crewai -o team.py

# LangGraph pipeline
agent-generator "ETL pipeline: extract, transform, load" -f langgraph -o pipeline.py

# WatsonX YAML agent
agent-generator "Customer support assistant" -f watsonx_orchestrate -o support.yaml

# Just test it (no credentials needed)
agent-generator "Hello world agent" -f crewai --dry-run

Step 3 -- Run

python team.py

Supported frameworks

Framework What you get
CrewAI crew.py + agents.yaml + tasks.yaml + tools + tests
LangGraph graph.py with StateGraph and typed state
WatsonX Orchestrate agent.yaml ready for orchestrate agents import
CrewAI Flow Flow class with @start / @listen
ReAct Reasoning loop with tool registry

Built-in tools

Your agents can use these out of the box:

Tool What it does
web_search Search the web
pdf_reader Read PDF files
http_client Call APIs
sql_query Query databases
file_writer Save files
vector_search Semantic search

Web UI

A visual wizard that guides you step by step:

uvicorn agent_generator.wsgi:app --port 8000

Open http://localhost:8000 -- describe your agents, pick a framework, download a ZIP.

Works with local LLMs (Ollama), remote LLMs (OllaBridge), or cloud APIs (OpenAI).

CLI options

agent-generator "your description" [options]
Flag What it does
-f crewai Pick framework: crewai, langgraph, watsonx_orchestrate, crewai_flow, react
-o file.py Save to file (otherwise prints to screen)
--dry-run Generate without calling any LLM
--mcp Add a FastAPI server wrapper
-p openai Use OpenAI instead of WatsonX
--show-cost Show estimated cost before generating

Docker

docker build -t agent-generator .
docker run -e WATSONX_API_KEY=... -p 8000:8000 agent-generator

Contributing

git clone https://github.com/ruslanmv/agent-generator.git
cd agent-generator
pip install -e ".[dev,all]"
pytest                # run tests
ruff check src/       # lint
mkdocs serve          # docs

License

Apache 2.0 -- PRs and issues welcome.

Yorumlar (0)

Sonuc bulunamadi