PROJECT KAIZEN

From Local Git Repos to Universal AI Context
A read-only architecture for portable AI context · Git stays the source of truth
1. SOURCE OF TRUTH
Local Git Forge
  • Infrastructure Repos
  • Application Repos
  • Service Docs
  • Runbooks
  • Skill Repos
  • Tooling Repos
  • Connector Repo
Read-only
Deploy Key
2. READ-ONLY INGEST
(Internal)

Packs repos & skills
into clean context

  • Manifest (files, sizes, tokens)
  • Include / Exclude Patterns
  • Token Protected Endpoints
/{token}/{owner}/{repo}
/{token}/{owner}/{repo}/manifest
OUTPUT
Packed Markdown
(Context)
Manifest
(JSON)
Internal
Token Auth
3. OAUTH MCP CONNECTOR
(Public Interface)

OAuth protected MCP server
exposes clean tools

  • list_repos
  • list_skills
  • search
  • fetch
  • get_repo_manifest
  • pack_skill
  • pack_repo_context
Curated Catalog (Aliases)
repo:project/main git:org/project
skill:infra-troubleshooting git:skills/infra
OAuth 2.0
(HTTPS)
4. AI CLIENTS / AGENTS
(Cloud & Local)
  • ChatGPT
  • Claude
  • Grok
  • Custom Agents
  • VS Code / Extensions
  • Local Models (UI)
Fresh Context
On Demand
5. OPERATOR LOOP
(You Control)
  • Scripts Repo (SSH)
  • Bootstrap Terminals
  • Install Helpers
  • Sync Skills & Tools
  • Keep Environments Consistent
SECURITY BY DESIGN
Cloud AI gets context, not credentials. Read-only by construction.
ONE SOURCE OF TRUTH
Git remains the authority. AI reads fresh context, not stale memory.
UNIVERSAL AI CONTEXT
Repos → Ingest → MCP → AI Clients. Fresh. Portable. Secure.
OPERATIONALLY USEFUL
AI becomes useful without becoming trusted. Reference, not authority.
STOP CARRYING CONTEXT
No copy-paste. No archaeology. Just current context, on demand.