Enterprise Patterns

Scale Wave across multiple projects and teams with shared configuration, centralized standards, and comprehensive audit trails.

Prerequisites

• Wave deployed to teams
• Git repository for shared configuration
• CI/CD pipeline access

Step-by-Step

1. Create Shared Configuration Repository

Centralize personas and contracts for consistency:

org-wave-config/
├── personas/
│   ├── navigator.md
│   ├── auditor.md
│   └── craftsman.md
├── contracts/
│   ├── gh-pr-review.schema.json
│   └── security-report.schema.json
└── README.md

2. Reference Shared Config

Projects include shared configuration via git submodule:

git submodule add git@github.com:org/org-wave-config.git .wave/shared

3. Override Per-Project

Each project extends shared configuration:

# wave.yaml
personas:
  # Use shared navigator
  navigator:
    adapter: claude
    system_prompt_file: .wave/shared/personas/navigator.md
    permissions:
      allowed_tools: [Read, Glob, Grep]
      deny: [Write, Edit, Bash]

  # Project-specific persona
  domain-expert:
    adapter: claude
    system_prompt_file: .wave/personas/domain-expert.md
    permissions:
      allowed_tools: [Read, Glob]
      deny: [Write, Edit, Bash]

Security Controls

Permission Enforcement

Personas define what AI can and cannot do:

personas:
  reader:
    permissions:
      allowed_tools: [Read, Glob, Grep]
      deny: [Write, Edit, Bash]

  writer:
    permissions:
      allowed_tools: [Read, Write, Edit]
      deny:
        - Bash(rm -rf *)
        - Write(/etc/*)

Audit Logging

Enable comprehensive logging for compliance:

runtime:
  audit:
    log_all_tool_calls: true
    log_all_file_operations: true
    log_dir: .wave/traces/

Logs capture:

• Tool calls made
• Files accessed
• Outputs generated
• Execution timestamps

API Key Management

Use environment variables, never files:

export ANTHROPIC_API_KEY=sk-...

In CI/CD, use secrets management:

# GitHub Actions
env:
  ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}

Scaling Patterns

Parallel Step Execution

Wave runs independent steps concurrently:

steps:
  - id: analyze
    # Runs first

  - id: security
    dependencies: [analyze]
    # Runs in parallel with quality

  - id: quality
    dependencies: [analyze]
    # Runs in parallel with security

  - id: summary
    dependencies: [security, quality]
    # Waits for both

Large Codebases

Limit scope with workspace mounts:

workspace:
  mount:
    - source: ./src/auth  # Only auth module
      target: /code
      mode: readonly

Timeout Configuration

Set appropriate limits:

runtime:
  default_timeout_minutes: 30

Complete Example

Enterprise-ready manifest:

apiVersion: v1
kind: WaveManifest
metadata:
  name: enterprise-project

adapters:
  claude:
    binary: claude
    mode: headless

personas:
  navigator:
    adapter: claude
    system_prompt_file: .wave/shared/personas/navigator.md
    permissions:
      allowed_tools: [Read, Glob, Grep]
      deny: [Write, Edit, Bash]

  auditor:
    adapter: claude
    system_prompt_file: .wave/shared/personas/auditor.md
    permissions:
      allowed_tools: [Read, Grep]
      deny: [Write, Edit, Bash]

  craftsman:
    adapter: claude
    system_prompt_file: .wave/personas/craftsman.md
    permissions:
      allowed_tools: [Read, Write, Edit]
      deny:
        - Bash(rm -rf *)
        - Write(/etc/*)

runtime:
  workspace_root: .wave/workspaces
  default_timeout_minutes: 30
  audit:
    log_all_tool_calls: true
    log_all_file_operations: true
    log_dir: .wave/traces/

Monitoring

Execution Status

# List runs
wave list

# Check specific run
wave status <run-id>

# View logs
wave logs <run-id>

Audit Trail

All executions logged to runtime.audit.log_dir:

.wave/traces/
├── run-abc123/
│   ├── step-analyze.log
│   └── step-review.log
└── run-def456/
    └── ...

Cost Management

Workspace Cleanup

Clean up old runs:

# Remove all workspaces
wave clean

# Keep last N runs
wave clean --keep 10

Best Practices

1. Centralize Standards - Shared personas and contracts ensure consistency
2. Start Small - Begin with one team, one pipeline; expand based on results
3. Review Changes - Pipeline configurations go through code review
4. Monitor Usage - Track which pipelines run, how often, and outcomes
5. Document Everything - Clear descriptions in pipeline metadata

Compliance

Wave's logging provides audit requirements:

• Tool calls made
• Files accessed
• Outputs generated
• Execution timestamps

Workspaces are ephemeral by default. Monitor disk usage and configure retention as needed.

Next Steps

• Audit Logging - Detailed audit configuration
• CI/CD Integration - Automate pipelines in builds
• Contracts - Validate pipeline outputs