AutoGen Adapter

Generate Python AutoGen agent code from your agent.yaml manifest.

Usage

export ANTHROPIC_API_KEY=your-api-key-here
agentspec generate agent.yaml --framework autogen --output ./generated/

Get an API key at console.anthropic.com.

Generated Files

File	Description
agent.py	AssistantAgent setup, run_agent() coroutine, CLI entry point
tools.py	Tool registry — imports and exposes AGENT_TOOLS list
tool_implementations.py	Typed async stub functions — fill in the bodies
manifest.py	Runtime manifest loader + get_capabilities()
agent.yaml	Copy of your manifest (ships with the agent)
server.py	FastAPI + SSE streaming server (when spec.api is set)
guardrails.py	Input/output guardrail functions (when spec.guardrails is set)
requirements.txt	Runtime dependencies
requirements-test.txt	Test dependencies (pytest)
.env.example	All required environment variables
docker-compose.yml	Agent + AgentSpec sidecar
README.md	Quick-start guide

Manifest Mapping

agent.yaml field	Generated code
spec.model.provider: openai	OpenAIChatCompletionClient from autogen_ext.models.openai
spec.model.provider: anthropic	AnthropicChatCompletionClient from autogen_ext.models.anthropic
spec.model.provider: azure	AzureOpenAIChatCompletionClient from autogen_ext.models.azure
spec.model.provider: groq	OpenAIChatCompletionClient with Groq base_url
spec.tools[]	Plain async functions in tool_implementations.py, registered in tools=[]
spec.memory	ListMemory (in-process) or Redis-backed ListMemory
spec.guardrails	run_input_guardrails() + run_output_guardrails() in guardrails.py
spec.prompts.system	system_message on AssistantAgent
spec.api	FastAPI /chat endpoint with SSE streaming via agent.run_stream()
spec.observability	Langfuse env vars wired before model client init

Example

metadata:
  name: research-assistant
  version: 1.0.0

spec:
  model:
    provider: openai
    id: gpt-4o
    apiKey: $env:OPENAI_API_KEY

  tools:
    - name: search-web
      description: Search the web for information
      annotations:
        readOnlyHint: true

  memory:
    shortTerm:
      backend: in-memory

  guardrails:
    input:
      - type: prompt-injection
      - type: pii-detector
        action: scrub

agentspec generate agent.yaml --framework autogen --output ./generated/
cd generated
pip install -r requirements.txt
cp .env.example .env
python agent.py "What are the latest AI research papers?"

See also

• LangGraph Adapter — Python, StateGraph-based
• CrewAI Adapter — Python, role-based crews
• Mastra Adapter — TypeScript
• Concepts: Adapters — how the adapter system works