07. 12-Factor Agents
What this is
12-factor-agents is best understood as a production engineering playbook for AI agents.
It is not mainly a framework that you import and build on directly.
It is closer to:
- a set of design principles
- a checklist for reliability
- a guide for building maintainable agent systems in the real world
Mental model
The analogy is:
12-factor app
↓
best practices for SaaS / cloud apps
12-factor agents
↓
best practices for production AI agents
So this topic belongs after you already understand:
- what an agent is
- how agent architectures work
- tool calling
- planning loops
- memory
- multi-agent coordination
Why it belongs later
If you study this too early, it may feel abstract.
That is because these principles make the most sense after you have already built or understood a few agent systems.
Then the questions become practical:
- How do I make this agent reliable?
- How do I manage tools safely?
- How do I observe what happened?
- How do I debug failures?
- How do I run this in production?
That is exactly where 12-factor ideas become valuable.
Best way to use the repo
Use the repo as:
- a reading reference
- a design checklist
- a review lens for your own agent projects
Do not treat it as “the agent framework you must adopt first”.
Suggested use in your roadmap
LLMs
↓
Tool use
↓
Agents
↓
Production agent design
↓
12-Factor Agents
That makes this the right final topic in the Agents module.
What to extract from it
When you study the repo, focus on questions like:
- What makes an agent system robust?
- What should be configurable?
- What should be observable?
- How should tools and prompts be managed?
- How do you reduce hidden state and brittle behavior?
Next
After this folder, return to the larger overview in 03-Advanced/13-Agents.md.