Global Configuration

Red Station configuration is centered in the ~/.pilot/ directory.

Directory Structure

~/.pilot/
├── config.yaml          # Main configuration overrides
├── agent-defaults.yaml  # Default settings for all agents
├── mcp.yaml             # MCP Server registry
├── models.yaml          # LLM Provider settings
├── tools.yaml           # Runner configuration
└── secrets.yaml         # Encrypted secrets store

agent-defaults.yaml

Defines the baseline behavior for any agent that doesn't explicitly override these values.

defaults:
  model:
    provider: ollama
    name: llama3.1:8b
    temperature: 0.7
    max_tokens: 4096
    
  behavior:
    autonomy: suggest    # Agents pause for approval by default
    max_iterations: 10   # Prevent infinite loops
    timeout: 300         # Seconds
    
  memory:
    type: conversation
    max_messages: 50
    
  security:
    sandbox: true        # Isolate execution
    allowed_paths:       # Whitelist accessible directories
      - "src/"
      - "tests/"
      - "docs/"
    denied_paths:
      - ".env"
      - "secrets/"

models.yaml

Configures the LLM backend.

default: ollama

providers:
  ollama:
    type: ollama
    base_url: http://localhost:11434
    
  openai:
    type: openai
    api_key: ${secrets.OPENAI_API_KEY}
    
aliases:
  fast: 
    provider: ollama
    model: llama3.1:8b
  smart:
    provider: openai
    model: gpt-4-turbo

mcp.yaml

Registers MCP Servers (the "Matrix").

servers:
  # Built-in Python implementation
  filesystem:
    transport: builtin
    enabled: true
    
  # External Node.js implementation
  github:
    command: npx
    args: ["-y", "@modelcontextprotocol/server-github"]
    env:
      GITHUB_TOKEN: "${secrets.GITHUB_TOKEN}"
    enabled: true

tools.yaml

Configures Runners (specialized command executors).

runners:
  test:
    type: test
    auto_detect: true  # Auto-finds pytest, jest, etc.
    coverage: true
    
  lint:
    type: lint
    auto_detect: true