ArchSpine Strategy

ArchSpine is positioned as a semantic control plane for AI-assisted software engineering. The goal is not to generate prettier docs; it is to make repository structure explicit, queryable, and governable.

The problem

Large repositories decay in predictable ways:

• God files absorb too much logic
• responsibilities blur across layers
• historical intent disappears as teams change

Traditional prompt-based AI workflows make this worse because they treat repository understanding as an ad hoc reconstruction problem.

The thesis

ArchSpine addresses that problem with three core ideas:

1. deterministic extraction
2. explicit governance
3. durable semantic memory

Deterministic extraction

Use AST-derived structure as the stable base so agents are not guessing at syntax or dependencies.

Governance

Let teams declare architectural rules in .spine/rules/, then audit and repair against those rules.

Semantic memory

Persist role, responsibility, and drift information so repository intent survives beyond individual contributors.

Execution model

The runtime must match the strategy:

• pipeline steps should use explicit stage input/output contracts
• shared runtime state should stay narrow and readable
• transient artifacts should be kept separate from telemetry
• orchestration should live in services, not leak into task internals

This keeps ArchSpine deterministic enough to govern, while still being practical for CLI-first workflows and future MCP or daemon-style entry points.

Open core boundary

Open-source layer:

• .spine protocol
• extractors
• base CLI
• local aggregation
• local MCP support

Commercial extensions, if pursued later, should focus on organization-scale control-plane value rather than basic repository generation.

Strategic moat

ArchSpine sits inside four high-value workflow moments:

1. commit-time sync and repository hygiene
2. CI and PR-time governance
3. agent context retrieval through MCP
4. onboarding and repository comprehension

Long-term goal

Make .spine feel as standard to AI-readable repositories as package.json feels to JavaScript projects.

Related docs

• Architecture Overview
• Task Execution Model