Core vs AI

Core vs AI

IntentFlow has two modes.

IntentFlow Core

Use Core when you want the architecture without extra AI contract files.

Core includes:

• Kotlin state, intent, event, effect, output, and route types
• pure reducer
• effect handler
• store
• projection
• tests

Core is good for:

• teams not using coding agents
• small packages
• manually written features
• gradual adoption inside existing apps

IntentFlow AI

Use AI mode when coding agents will create, edit, or review features.

AI mode includes everything in Core plus:

• .intentflow.yaml manifest
• invariants
• acceptance traces
• generator validation
• provider-specific AI instructions
• compact intentflow ai-context output

AI mode is good for:

• Codex implementation loops
• Claude Code feature edits
• Gemini architecture review
• Copilot inline help and PR review
• Cursor editor-native refactors
• teams that want AI to obey explicit rules

Difference In One Table

Concern	Core	AI
Source of truth	Kotlin types and tests	Kotlin types, tests, and manifest
AI safety	Human review	Human review plus machine-readable contract
Generated files	Kotlin skeleton	Kotlin skeleton plus manifest
Context budget	Manual judgment	Compact agent context command
Drift detection	Tests and code review	Tests, manifest validation, acceptance traces

When To Start With Core

Start with Core if:

• the feature contract is still unclear
• the team is learning the pattern
• the feature is small and manually maintained
• you do not want manifest overhead yet

You can add AI mode later.

When To Start With AI

Start with AI if:

• a coding agent will implement or modify the feature
• the feature has many states
• workflow rules must be stable over time
• reviewers need compact acceptance traces
• you want generated code to stay inside architectural boundaries

Rule Of Thumb

Core is the architecture.

AI mode is the architecture plus a contract that humans and machines can share.