Skip to content
Workflow MCP

Introduction

Workflow MCP Server

Section titled “Workflow MCP Server”

Workflow MCP is a Model Context Protocol (MCP) server that provides structured workflow automation for AI-assisted development with Claude.

What is MCP?

Section titled “What is MCP?”

The Model Context Protocol is an open standard for connecting AI assistants to external tools and data sources. MCP servers expose tools that AI models can invoke to perform actions.

What Does Workflow MCP Do?

Section titled “What Does Workflow MCP Do?”

Workflow MCP provides tools for:

Session Management

Start, pause, and resume workflow sessions with automatic state persistence.

Structured Logging

Log findings, decisions, tasks, and issues with automatic Entry-ID generation.

Workflow Commands

Support for /research, /plan, /implement, /review, /debug, and more.

Handoff Generation

Generate handoff messages for transitioning between workflow phases.

Key Features

Section titled “Key Features”

Dynamic Problem Solving (DPS)

Section titled “Dynamic Problem Solving (DPS)”

Control how strictly Claude follows your workflow with DPS levels (0-5):

  • DPS 0: Strict role adherence - stops on any deviation
  • DPS 1-2: Minor adjustments allowed
  • DPS 3-4: Can propose alternatives and redesigns
  • DPS 5: Maximum flexibility

Automatic Input Linking

Section titled “Automatic Input Linking”

Each workflow phase automatically finds relevant files from previous phases:

  • /plan reads from research files
  • /implement reads from plan and research files
  • /review reads from all previous phases

Iteration Tracking

Section titled “Iteration Tracking”

When multiple sessions are started for the same topic on the same day, the system:

  • Detects this is iteration #2, #3, etc.
  • Prompts for reflection on why multiple sessions are needed
  • Helps identify scope issues or process improvements

Local LLM Delegation

Section titled “Local LLM Delegation”

Offload mechanical coding tasks to local models (e.g., DeepSeek via Ollama):

  • Preserve Claude’s context window for reasoning
  • Generate boilerplate, tests, and documentation locally
  • Automatic task scoring to suggest what to delegate

Dual Transport Support

Section titled “Dual Transport Support”
  • stdio: For local CLI usage with Claude Code or any MCP client
  • HTTP: For remote deployment with OAuth authentication

Portable Architecture

Section titled “Portable Architecture”
  • Works across projects and machines
  • Session state persists in JSON files
  • Configure via environment variables

Why Use Workflow MCP?

Section titled “Why Use Workflow MCP?”
  1. Structure: Brings methodology to AI-assisted development
  2. Traceability: Every decision, finding, and blocker is logged with Entry-IDs
  3. Continuity: Sessions can be paused and resumed with full context
  4. Flexibility: Adapts to your workflow with DPS levels
  5. Efficiency: Delegate mechanical work to local LLMs
  6. Integration: Works with Claude Code, Claude Desktop, and any MCP-compatible client

Next Steps

Section titled “Next Steps”