12-factor-agentops
factorsbodenfuller.comgithub
/factors/bodenfuller.com/github
← back

The 12 Factors

Battle-tested patterns for reliable AI agent operations. Start with Factor I and progress through each tier.

I
foundation

Automated Tracking

Track everything in git. Learn from history.

3 min read
read
II
foundation

Context Loading

Load when needed, not all at once. Stay under 40%.

3 min read
read
III
foundation

Focused Agents

One agent, one job. Compose from simple parts.

5 min read
read
IV
foundation

Continuous Validation

Validate at every step. Catch errors in seconds.

6 min read
read
V
operations

Measure Everything

Track the five core metrics. Know when things break.

7 min read
read
VI
operations

Resume Work

Resume across days without context loss.

5 min read
read
VII
operations

Smart Routing

Right task, right agent, every time.

4 min read
read
VIII
operations

Human Validation

Humans in the loop for critical decisions.

5 min read
read
IX
improvement

Mine Patterns

Capture what works. Build institutional memory.

5 min read
read
X
improvement

Small Iterations

Always improving. Never stagnant.

5 min read
read
XI
improvement

Fail-Safe Checks

Clear boundaries. Safe automation.

3 min read
read
XII
improvement

Package Patterns

Package once, deploy everywhere.

3 min read
read
> making AI agents reliable
githubnpmby bo