Skip to content

1-preliminaries.md

Latest commit

 

History

History
92 lines (60 loc) · 3.62 KB

1-preliminaries.md

File metadata and controls

92 lines (60 loc) · 3.62 KB
Raw

MCP Setup Guide: Preliminaries

Quick Overview

What: Model Context Protocol (MCP) servers provide AI assistants with secure, standardized access to local resources (files, databases, APIs).

Why: Extends your existing ./ai/ documentation framework with dynamic capabilities - AI can now execute actions, not just read context.

Security: MCP adds new attack vectors. Solution: restrict access to non-sensitive paths only (data-public/, analysis/, ai/). Keep data-private/ excluded.

Enterprise Challenge: Node.js installed but not in PATH. Solution: session-based setup scripts + VS Code tasks.

Status: ✅ Node.js v22.12.0 installed, ✅ automated setup system created, ⏳ MCP servers pending.

Next: Install file system MCP server with restricted permissions for safe experimentation.

Detailed Implementation

Problem Context

  • Current System: Human-curated ./ai/ files provide AI with project context and methodology
  • Limitation: AI can only read what humans manually share - no direct file/data access
  • Goal: Add dynamic capabilities while maintaining security and epistemological control

MCP Architecture

AI Assistant ↔ VS Code (MCP Client) ↔ MCP Server ↔ Local Resources

Key Insight: MCP provides "hands" (execution) to complement ./ai/ "brain" (knowledge framework).

Security Analysis

New Risk Vectors: - Direct file system access beyond curated content - Potential data exfiltration to external AI services - Privilege escalation through user permissions

Mitigation Strategy:

HIGH SENSITIVITY → Manual curation via ./ai/ system (current approach)
LOW SENSITIVITY → MCP access (new capability)

Recommended Access Pattern

  • ✅ ./data-public/ - Safe for MCP
  • ✅ ./analysis/ - Safe for MCP
  • ✅ ./ai/ - Safe for MCP
  • ❌ ./data-private/ - Manual curation only

Enterprise Installation Challenges

Issue: Node.js v22.12.0 installed by enterprise but not added to system PATH.

Solution: Session-based setup with automation:

  1. Smart Setup Script (setup-nodejs.ps1):
    • Checks if Node.js already available in session
    • Adds to PATH only if needed
    • Provides status feedback
  2. Automated Project Status (project-status.ps1):
    • Automatically ensures Node.js setup
    • Validates project structure
    • Checks MCP server status
    • Reports git status
  3. VS Code Integration:
    • Tasks for one-click execution
    • Keyboard shortcuts (Ctrl+Shift+S for status)
    • Automatic environment preparation

Technical Implementation

Files Created: - setup-nodejs.ps1 - Smart Node.js environment setup - project-status.ps1 - Comprehensive project status with auto-setup - .vscode/tasks.json - VS Code task definitions - .vscode/keybindings.json - Keyboard shortcuts

Usage Pattern:

# Manual (any session)
.\setup-nodejs.ps1

# Automated (VS Code)
Ctrl+Shift+S  # Full project status + Node.js setup
Ctrl+Shift+N  # Node.js setup only

Next Phase: MCP Server Installation

Planned Approach: 1. Install file system MCP server with restricted permissions 2. Configure access limited to safe directories 3. Test with non-sensitive analysis workflows 4. Validate security boundaries 5. Document integration with existing ./ai/ framework

Success Criteria: - AI can access analysis code and public data - Private data remains protected - Existing methodology documentation continues to guide AI behavior - No degradation of epistemological control