Migration Guides -- AgentBreeder

Bring your existing AI agents to AgentBreeder in minutes. Your framework code stays the same. AgentBreeder wraps it with governance, multi-cloud deploy, and org-wide discoverability.

Why Migrate?

You already built an agent. It works. But now you need:

What AgentBreeder does NOT do: It does not replace your framework. Your LangGraph graphs, CrewAI crews, OpenAI agents, and custom code run exactly as-is inside the container AgentBreeder builds. Think of it as the deployment and governance layer that sits around your agent.

Quick Decision Matrix

5-Minute Quick Start (Any Framework)

Regardless of your framework, the migration follows the same three steps:

Step 1: Keep your agent code

my-agent/
  agent.py          # <-- your existing code, unchanged
  requirements.txt  # <-- your existing deps

Step 2: Add agent.yaml

name: my-agent
version: 1.0.0
team: my-team
owner: me@company.com
framework: langgraph  # or crewai, openai_agents, custom

model:
  primary: gpt-4o

deploy:
  cloud: local

Step 3: Deploy

pip install agentbreeder
agentbreeder deploy agent.yaml

That is the entire migration. Everything else -- RBAC, registry, cost tracking, health checks -- happens automatically.

Feature Comparison: What AgentBreeder Adds

What Stays the Same After Migration

This is important: AgentBreeder does not modify your agent code. Your framework-specific logic runs inside a container that AG builds and manages. Specifically:

• Your agent.py / main.py files are unchanged
• Your requirements.txt / pyproject.toml are preserved (AG adds framework deps if missing)
• Your LLM API calls go through the same providers
• Your tools, prompts, and logic are identical
• You can still run your agent locally without AgentBreeder (just python agent.py)

What AG adds is a server wrapper (server.py) that exposes your agent as an HTTP service with /invoke and /health endpoints, plus the deploy infrastructure.

Architecture: How It Works

                    Your Code              AgentBreeder Adds
                 +--------------+     +------------------------+
                 |  agent.py    |     |  server.py (wrapper)   |
                 |  tools.py    |     |  Dockerfile (generated)|
  agent.yaml -> |  prompts/    | --> |  Health checks         |
                 |  requirements|     |  OpenTelemetry sidecar |
                 +--------------+     +------------------------+
                                              |
                                    +---------+---------+
                                    |                   |
                              agentbreeder deploy        agentbreeder deploy
                              --target local       --target aws
                                    |                   |
                              Docker Compose       ECS Fargate
                                    |                   |
                              localhost:8080     https://my-agent.ecs.aws
                                    |                   |
                              +-----+-------------------+-----+
                              |         Registry              |
                              |   RBAC | Audit | Cost Track   |
                              +-------------------------------+

Common Questions

Q: Do I need to rewrite my agent? No. Your agent code is unchanged. You add an agent.yaml file and run agentbreeder deploy.

Q: Can I still run my agent without AgentBreeder? Yes. Your agent.py still works standalone. AgentBreeder is additive.

Q: What if my framework isn't listed? Use framework: custom. See FROM_CUSTOM.md. Any Python agent that can be called via a function works.

Q: Does AgentBreeder lock me in? No. Your agent code has zero AgentBreeder imports. The agent.yaml is a declarative config file. You can stop using AG at any time and deploy your container manually.

Q: What about multi-agent systems? AgentBreeder has orchestration.yaml for defining multi-agent workflows (sequential, parallel, router, supervisor, fan-out/fan-in). Your framework's native orchestration (LangGraph subgraphs, CrewAI processes, OpenAI handoffs) also works as-is inside the container.

Next Steps

Pick your migration guide:

1. Migrate from LangGraph
2. Migrate from CrewAI
3. Migrate from OpenAI Agents SDK
4. Migrate from Microsoft AutoGen
5. Bring Your Own Agent (Custom)