cortex-tms

Documentation Governance for AI Coding Agents

⭐ 166+ GitHub Stars | Open source, community-driven

---

What is Cortex TMS?

Cortex TMS scaffolds and validates governance documentation for AI coding agents. As AI models get more powerful and autonomous, they need clear, current governance docs to stay aligned with your project standards.

The Challenge: Modern AI agents handle large context windows and can work autonomously—but without governance, they drift from your standards, overengineer solutions, and write inconsistent code.

The Solution: Cortex TMS provides:

• 📋 Documentation scaffolding - Templates for PATTERNS.md, ARCHITECTURE.md, CLAUDE.md
• ✅ Staleness detection - Detects when governance docs go stale relative to code changes (v4.0)
• 🔍 Structure validation - Automated health checks in CI or locally
• 📦 Archive management - Keep task lists focused and maintainable

---

Three Pillars

1. 📋 Consistency - Document Your Standards

Scaffold governance docs that AI agents actually read:

• PATTERNS.md - Code patterns and conventions
• CLAUDE.md - Agent workflow rules (git protocol, scope discipline, human approval gates)
• ARCHITECTURE.md - System design and tech stack
• DOMAIN-LOGIC.md - Business rules and constraints

Result: AI writes code that follows YOUR patterns, not random conventions from its training data.

2. 🔍 Freshness - Detect Staleness

New in v4.0: Git-based staleness detection catches when docs go stale:

cortex-tms validate

⚠️  Doc Staleness
    PATTERNS.md may be outdated
    Doc is 45 days older than code with 12 meaningful commits
    Code: 2026-02-20
    Doc:  2026-01-06

    Review docs/core/PATTERNS.md to ensure it reflects current codebase

How it works: Compares doc modification dates vs code commit activity. Flags stale docs before they mislead AI agents.

Note: Staleness v1 uses git timestamps (temporal comparison only). Cannot detect semantic misalignment. Future versions will add semantic analysis.

3. 🛡️ Safety - Human Oversight

CLAUDE.md governance rules require human approval for critical operations:

• Git commits/pushes require approval
• Scope discipline prevents overengineering
• Pattern adherence enforced through validation

Result: AI agents stay powerful but don't run wild.

---

What Cortex Does (and Doesn't Do)

✅ What Cortex Does

• Scaffolds governance docs - Templates for common project documentation
• Validates doc health - Checks structure, freshness, completeness
• Detects staleness - Flags when docs are outdated relative to code (v4.0)
• Enforces size limits - Keeps docs focused and scannable
• Archives completed tasks - Maintains clean NEXT-TASKS.md
• Works in CI/CD - GitHub Actions validation templates included

❌ What Cortex Does NOT Do

• Not a token optimizer - Validates documentation health, not context size
• Not code enforcement - Validates DOCUMENTATION health, not code directly
• Not a replacement for code review - Complements human review, doesn't replace it
• Not semantic analysis (yet) - Staleness v1 uses timestamps, not AI-powered diff analysis

---

Quick Start

# Initialize governance docs in your project
npx cortex-tms@latest init

# Validate doc health (including staleness detection)
npx cortex-tms@latest validate

# Strict mode (warnings = errors, for CI)
npx cortex-tms@latest validate --strict

# Check project status
npx cortex-tms@latest status

# Archive completed tasks
npx cortex-tms@latest archive --dry-run

Installation: No installation required with npx. For frequent use: npm install -g cortex-tms@latest

---

CLI Commands

cortex-tms init

Scaffold TMS documentation structure with interactive scope selection.

cortex-tms init                          # Interactive mode
cortex-tms init --scope standard         # Non-interactive
cortex-tms init --preset node            # Node.js governance pack
cortex-tms init --preset python          # Python governance pack
cortex-tms init --preset go              # Go governance pack
cortex-tms init --preset node --scope standard --force  # CI-friendly
cortex-tms init --dry-run                # Preview changes

Governance packs (--preset) fill template content with ecosystem-specific patterns, conventions, and idioms. Independent of --scope (which controls the file set). Available presets: node, python, go.

cortex-tms validate

Verify project TMS health with automated checks.

cortex-tms validate         # Check project health
cortex-tms validate --fix   # Auto-repair missing files
cortex-tms validate --strict # Strict mode (warnings = errors)

What it checks:

• ✅ Mandatory files exist (NEXT-TASKS.md, CLAUDE.md, copilot-instructions.md)
• ✅ File size limits (Rule 4: HOT files < 200 lines)
• ✅ Placeholder completion (no [Project Name] markers left)
• ✅ Archive status (completed tasks should be archived)
• ✅ Doc staleness (NEW in v4.0) - governance docs current with code

cortex-tms status

Text summary of project health and sprint progress.

cortex-tms status  # Health summary with progress bars

Shows: project identity, validation status, sprint progress, backlog size.

cortex-tms dashboard ✨ New in v4.0

Full-screen interactive terminal UI for governance health monitoring.

cortex-tms dashboard        # Interactive dashboard (navigate with 1/2/3 keys)
cortex-tms dashboard --live # Auto-refresh every 5 seconds

Three views (switch with number keys):

• 1 — Overview: Governance health score (0–100), staleness status, sprint progress
• 2 — Files: HOT files list, HOT/WARM/COLD distribution, file size health
• 3 — Health: Validation status, Guardian violation summary

cortex-tms archive

Archive completed tasks and old content.

cortex-tms archive           # Archive completed tasks
cortex-tms archive --dry-run # Preview what would be archived

Archives completed tasks from NEXT-TASKS.md to docs/archive/ with timestamp.

Note: cortex-tms auto-tier is deprecated—use archive instead.

cortex-tms migrate

Intelligent version management—detect outdated templates and upgrade safely.

cortex-tms migrate                    # Analyze version status
cortex-tms migrate --apply            # Auto-upgrade OUTDATED files
cortex-tms migrate --rollback         # Restore from backup

cortex-tms prompt

Access project-aware AI prompts from the Essential 7 library.

cortex-tms prompt              # Interactive selection
cortex-tms prompt init-session # Auto-copies to clipboard

cortex-tms review 🛡️

Guardian: AI-powered semantic validation against project patterns.

cortex-tms review src/index.ts              # Validate against PATTERNS.md
cortex-tms review src/index.ts --safe       # High-confidence violations only

cortex-tms tutorial

Interactive walkthrough teaching the Cortex Way.

cortex-tms tutorial  # 5-lesson guided tour (~15 minutes)

cortex-tms hooks

Manage git hooks for automatic documentation validation. Installs a pre-commit hook that runs cortex-tms validate before every commit.

cortex-tms hooks install              # Install pre-commit hook (default mode)
cortex-tms hooks install --strict     # Warnings also block commits
cortex-tms hooks install --skip-staleness  # Skip staleness checks (faster)
cortex-tms hooks status               # Show current hook configuration
cortex-tms hooks uninstall            # Remove the hook

Safety: Never overwrites foreign hooks. Only manages hooks with its own marker. Requires .cortexrc (run cortex-tms init first).

cortex-tms mcp 🔌

Start a read-only MCP (Model Context Protocol) server that exposes your TMS governance docs as resources to any AI tool that speaks MCP — Claude Desktop, Cursor, Windsurf, and future clients.

cortex-tms mcp                  # Start MCP server (STDIO transport)
cortex-tms mcp --print-config   # Print paste-ready client config snippets

Resources exposed (only files that exist on disk are advertised):

URI	File	Description
cortex://next-tasks	NEXT-TASKS.md	Active tasks and sprint backlog
cortex://claude	CLAUDE.md	AI agent instructions and workflow
cortex://patterns	docs/core/PATTERNS.md	Code patterns and conventions
cortex://architecture	docs/core/ARCHITECTURE.md	System architecture
cortex://domain-logic	docs/core/DOMAIN-LOGIC.md	Business logic and domain model
cortex://decisions	docs/core/DECISIONS.md	Architecture decision records
cortex://troubleshooting	docs/core/TROUBLESHOOTING.md	Common issues and solutions
cortex://glossary	docs/core/GLOSSARY.md	Terminology and definitions
cortex://schema	docs/core/SCHEMA.md	Data schemas and API contracts
cortex://copilot-instructions	.github/copilot-instructions.md	Copilot instructions
cortex://prompts	PROMPTS.md	Reusable prompt templates
cortex://agents	AGENTS.md	Multi-agent governance registry
cortex://future-enhancements	FUTURE-ENHANCEMENTS.md	Roadmap and backlog

Client setup — run cortex-tms mcp --print-config for paste-ready snippets. Quick reference:

Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "cortex-tms": {
      "command": "npx",
      "args": ["-y", "cortex-tms", "mcp"],
      "cwd": "/path/to/your/project"
    }
  }
}

Notes:

• Read-only — no write tools exposed. URI slugs are stable across projects.
• Respects .cortexrc custom paths (paths.docs, paths.tasks).
• Requires .cortexrc (run cortex-tms init first).

---

Documentation Structure

Folder / File	Purpose	Tier
NEXT-TASKS.md	Active sprint and current focus	HOT (Always Read)
PROMPTS.md	AI interaction templates (Essential 7)	HOT (Always Read)
CLAUDE.md	CLI commands & workflow config	HOT (Always Read)
.github/copilot-instructions.md	Global guardrails and critical rules	HOT (Always Read)
FUTURE-ENHANCEMENTS.md	Living backlog (not current sprint)	PLANNING
docs/core/ARCHITECTURE.md	System design & tech stack	WARM (Read on Demand)
docs/core/PATTERNS.md	Canonical code examples (Do/Don't)	WARM (Read on Demand)
docs/core/DOMAIN-LOGIC.md	Immutable project rules	WARM (Read on Demand)
docs/core/GIT-STANDARDS.md	Git & PM conventions	WARM (Read on Demand)