ADVANCED · 8 MODULES · 64 LESSONS
AI Code Governance and Agent Architecture
Set rules that neither people nor coding agents can skip, so your team gains speed with AI without losing control of the architecture.
- For you if
- You lead a team where coding agents already open pull requests and you worry about losing direction.
- By the end
- You'll know how to block architecture-breaking changes in CI, limit what each agent can touch, and require human approval where risk is high.
- You need
- Experience on backend teams with CI/CD in production, and Python to run the simulations.

Curriculum
02Fitness functions como gobernanza
03El modelo de permisos para agentes de código
- Module introduction: the permission model for code agents
- Least privilege for code agents
- Capabilities, not a single role
- The blast radius of an autonomous agent
- Read, write, and deploy as distinct levels
- The sandbox: isolating where the agent runs
- Why broad permissions are an enormous risk (and why the permission model is architecture)
- Project: design the permission model of Mercado's patch agent
05Procedencia y rendición de cuentas
- Module introduction: provenance and accountability
- Tracking what AI generated
- The audit trail
- CODEOWNERS and accountability
- Review coverage as a metric
- Hallucinated dependencies and slopsquatting
- License and security review of the generated code
- Project: provenance and audit of the agent contributions to Mercado
06La arquitectura como guardarrail del agente
- Module introduction: architecture as the agent's guardrail
- The architecture constitution
- ADRs as agent-readable constraints
- Giving the agent the right architecture context
- Policy-as-code: the rule the agent can't skip
- Keeping the context and the guardrail in sync
- Mercado's constitution
- Project: codify Mercado's architecture as a guardrail
08Proyecto — gobierna las contribuciones de código IA a Mercado
Reviews
What students say
These reviews are from enrolled students who completed at least 50% of the course. We moderate reviews only on content grounds (spam, offensive language, personal data), never for being critical or negative.
No approved reviews yet.
Be the first to share your experience!