-
Notifications
You must be signed in to change notification settings - Fork 21
Integration Guides
github-actions[bot] edited this page May 6, 2026
·
4 revisions
MCP Task Orchestrator can be integrated at different levels of sophistication. These guides are progressive — each tier builds on the previous. Start where you are, level up when you need more.
| Tier | Guide | Audience | What You Get |
|---|---|---|---|
| 1 | Bare MCP | Any MCP client user | 13 tools, persistent SQLite, role-based workflow |
| 2 | CLAUDE.md-Driven | Claude Code users | Consistent agent behavior via project instructions |
| 3 | Note Schemas | Schema configurers | Phase gate enforcement, required documentation |
| 4 | Plugin: Skills and Hooks | Claude Code + plugin | Automated workflows, plan-mode pipeline, subagent protocols |
| 5 | Output Styles | Power users | Full orchestrator mode with delegation |
| 6 | Self-Improving Workflow | Self-optimizers | Feedback loop, observation logging, auto-memory correction |
- "I use a non-Claude-Code MCP client" → Tier 1
- "I use Claude Code but don't want a plugin" → Tiers 1-2
- "I want to enforce documentation quality at each phase" → Add Tier 3 to any tier
- "I want automated skill workflows and plan-mode integration" → Tier 4
- "I want Claude to orchestrate and delegate, not implement directly" → Tier 5
- "I want the system to improve itself over time" → Tier 6
Each tier adds on top of the previous. Note schemas (Tier 3) can be added independently at any tier — they require only a .taskorchestrator/config.yaml file and no plugin or CLAUDE.md changes. The self-improving workflow (Tier 6) is also relatively standalone — it can be adapted with or without the full output style.
Getting Started
Integration Guides
- Overview
- Bare MCP
- CLAUDE.md-Driven
- Note Schemas
- Plugin: Skills & Hooks
- Output Styles
- Self-Improving Workflow
Reference
Operations
Project