🛠️ agent‑generator

A one‑line prompt … a fully configured multi‑agent team, served via an MCP Gateway.

agent‑generator converts plain‑English requirements into runnable code or YAML workflows for:

• CrewAI & CrewAI Flow
• LangGraph
• ReAct
• WatsonX Orchestrate (YAML)

You can choose your LLM provider:

• IBM WatsonX (default)
• OpenAI (via --provider openai or AGENTGEN_PROVIDER=openai)

---

Why Multi‑Agent + MCP Gateway?

Splitting a complex task into specialized agents lets you:

• Parallelize subtasks (e.g. research, summarise, review)
• Specialize prompts per role for cleaner, maintainable code
• Visualize your workflow as Mermaid/DOT diagrams

The MCP server wrapper (FastAPI) and MCP Gateway provide:

• Unified HTTP entrypoint for all agents
• Route isolation per task or agent
• Built‑in monitoring, health checks, and cost endpoints

This makes your multi‑agent system production‑ready with minimal boilerplate.

---

Highlights

Feature	✓
IBM WatsonX default	✅
OpenAI support	✅
Plug‑in providers	✅
MCP server wrapper & Gateway	✅
Rich CLI + Flask UI	✅
Mermaid / DOT diagrams	✅

---

Quick install

pip install agent-generator[web,openai]

Jump in: Installation ➜ · Usage ➜ · Frameworks ➜