How to decouple the existing architecture into a Distributed Multi-Agent System?Sub-agents vs. tools
![Gatekeeper root agent with tools=[verify_user_id_tool] and sub_agents=[registratore_agent, cameriere_agent]](_shared/img/subagents-code.png)
🛠️ Tools
rigid functions written in pure code (e.g. GPS calculation or saving a user to a database).
🤖 Sub-agents
secondary AI agents managed by a primary agent to execute specific tasks.
But what if…
Integration
What if we need to interact with components written in a different language?
Maintenance
What if certain components of our system need frequent updates and maintenance?
Scalability
What if the different components should scale at different rates?
Security
What if one of our components is compromised?