Skip to main content

Claude's Global Scope — The CEO Layer

Claude operates as a global orchestrator across all repositories. This isn't per-project tooling — it's a persistent identity with memory, preferences, and strategic context that spans everything.

How Claude's Brain Works​

The Three Layers​

1. Global Identity (~/.claude/CLAUDE.md)​

Loaded in every conversation, regardless of project. Contains:

  • Claude's role as CEO/Orchestrator
  • Ryan's preferences (concise, ship simple, no over-engineering)
  • The Pulp Fiction workflow
  • Secret management rules (direnv + age)
  • MCP server configuration

2. Persistent Memory (~/.claude/projects/*/memory/)​

Per-project learnings that survive across sessions:

  • Patterns and conventions confirmed over time
  • Key architectural decisions
  • Solutions to recurring problems
  • Project-specific agent configurations

3. Project Context (CLAUDE.md per repo)​

Repo-specific instructions and constraints:

  • Tech stack and dependencies
  • Testing patterns
  • Deployment procedures
  • Team-specific conventions

Claude's Responsibilities​

As CEO/Orchestrator, Claude:

  1. Receives direction from Ryan — understands intent, asks clarifying questions
  2. Delegates to specialists — spins up BMAD agents for planning, OpenClaw bots for execution
  3. Maintains strategic context — remembers decisions, patterns, and preferences across sessions
  4. Ensures quality — reviews before shipping, catches inconsistencies
  5. Documents decisions — pushes knowledge to the wiki as the single source of truth
  6. Manages the team — knows Max and Hank's capabilities, assigns work accordingly

Relationship to OpenClaw Bots​

Max and Hank are hands, not brains. They execute tasks assigned through GitHub but don't make strategic decisions. Claude defines what needs to happen; they make it happen.