Skip to main content
Glama
Workflow MCP Server

Workflow MCP Server

by tuanknguyen

Workflow MCP Server

A Python MCP server that guides agents through structured workflows. This server ensures agents follow predefined steps while maintaining flexibility in execution.

Development Approach

Package Management with UV

This project exclusively uses uv for all package management operations. Always use uv commands for:

  • Virtual environment creation: uv venv

  • Package installation: uv pip install

  • Running Python scripts: uv run python script.py

  • Running tests: uv run pytest

  • Running the application: uv run workflow-mcp

UV provides faster, more reliable package management than traditional tools. If you don't have uv installed, follow instructions at https://github.com/astral-sh/uv

Related MCP server: sop-mcp

Installation

Setting up the environment

# Create a virtual environment
uv venv

# Activate the virtual environment
source .venv/bin/activate  # Linux/macOS
# OR
.venv\Scripts\activate     # Windows

# Install the package
uv pip install -e .

# For development with testing tools
uv pip install -e ".[dev]"

Usage

Running the server

# Using the entry point script
uv run workflow-mcp

# Or directly with module
uv run python -m workflow_mcp_server

With SSE instead of stdio:

uv run workflow-mcp --sse --port 8888

Running tests

uv run pytest

Workflow Definition

Create YAML files in the frameworks directory with the following structure:

name: "Simple Workflow"
description: "A linear workflow with basic steps"
version: "1.0"
steps:
  - id: "step1"
    type: "instruction"
    content: "This is what you need to do first"
    next: "step2"
    
  - id: "step2"
    type: "tool_call"
    tool: "tool_name"
    parameters:
      param1: "value1"
    next: "step3"
    
  - id: "step3"
    type: "end"
    content: "Workflow complete"

Available Tools

  • list_workflows(): Lists available workflow frameworks

  • start_workflow(workflow_id): Start a new workflow session

  • complete_step(session_id, result): Mark current step as complete and get the next step

License

MIT

Available Tools

8 tools
complete_stepB

Complete current step and get next step.

Args: ctx: MCP context for logging and error handling session_id: ID of the session. result: Optional result from executing the current step.

Returns: Dictionary containing next step information.

Inputs: session_id (required), result. Returns: result

create_frameworkB

Create a new workflow framework.

Args: ctx: MCP context for logging and error handling schema: The framework schema. name: Optional name to use for the framework file.

Returns: The framework ID.

Inputs: schema (required), name. Returns: result

delete_sessionB

Delete a workflow session.

Args: ctx: MCP context for logging and error handling session_id: ID of the session to delete.

Returns: True if the session was deleted, False if it didn't exist.

Inputs: session_id (required). Returns: result

execute_current_stepC

Execute the current step automatically.

Args: ctx: MCP context for logging and error handling session_id: ID of the session.

Returns: Dictionary containing next step information.

Inputs: session_id (required). Returns: result

get_session_infoC

Get information about a workflow session.

Args: ctx: MCP context for logging and error handling session_id: ID of the session.

Returns: Dictionary containing session information.

Inputs: session_id (required). Returns: result

list_sessionsB

List all available workflow sessions.

Returns: List of session metadata.

Inputs: no named parameters. Returns: result

list_workflowsA

List all available workflows.

Returns: List of available workflow metadata.

Inputs: no named parameters. Returns: result

start_workflowA

Start a new workflow.

Args: ctx: MCP context for logging and error handling workflow_id: ID of the workflow to start.

Returns: Dictionary containing session ID and first step.

Inputs: workflow_id (required). Returns: result

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    B3.2/5.0

    Scored across 8 tools

    Disambiguation4/5

    Tools are largely distinct, focusing on framework creation, session management, and step execution. The only potential confusion is between complete_step and execute_current_step, but their descriptions clarify different actions: completing vs executing steps.

    Naming Consistency5/5

    All tool names follow a consistent verb_noun pattern in snake_case (e.g., create_framework, list_sessions, complete_step). The pattern is uniform and predictable.

    Tool Count5/5

    With 8 tools, the server covers core workflow operations without being excessive. The number is appropriate for a workflow management server's typical needs.

    Completeness3/5

    The tool set covers basic lifecycle (create framework, start, execute steps, manage sessions), but lacks tools for editing workflow definitions, listing steps, or pausing/aborting sessions, which are notable gaps.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      A
      quality
      D
      maintenance
      Provides LLMs with structured workflow guidance and next-step instructions for development procedures, ensuring consistent practices across projects.
      5
      24 PyPI
      1
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      Enables AI agents to execute multi-step Standard Operating Procedures step by step, with enforcement of completion at each step, making LLM behavior predictable and auditable.
      5
      3
      Apache 2.0
    • A
      license
      Not graded
      quality
      D
      maintenance
      Enables agents to navigate users through pre-defined decision trees by submitting answer labels, returning deterministic next instructions without extra LLM calls. Supports domain-independent procedural workflows with state tracking, handling of unknown responses, and a read-only Web UI.
      MIT