Workflows MCP
by AgentDeskAI
Workflows MCP is a Model Context Protocol (MCP) server that allows you to orchestrate and combine prompts and MCP servers into compound prompting tools. It acts like a dynamic prompting library that can be shared and version controlled through YAML files, defining how to use many MCP tools across MCP servers for specific tasks.
Last updated: N/A
What is Workflows MCP?
Workflows MCP is a server designed to simplify and enhance the use of Model Context Protocol (MCP) by allowing users to create and manage complex workflows involving multiple prompts and MCP servers. It turns prompts into programmable tools and rules for AI in a code editor, aiming to make AI interactions more deterministic and reduce overall token usage.
How to use Workflows MCP?
To use Workflows MCP, install the server into an MCP client using the provided npx command or JSON configuration. Configure workflows by creating YAML files in a .workflows or .mcp-workflows directory. Customize workflows with prompts, tool sequences, and parameter injections. Trigger workflows using custom commands within your MCP client.
Key features of Workflows MCP
Combine prompts with MCP servers into reusable workflows
Trigger workflows with custom commands
Define custom strategies for tool usage (sequential or situational)
Easily onboard teams with version-controlled prompts
Custom configuration via YAML files
Ready-to-use presets for thinking, coding, and GitHub tasks
Parameter injection for dynamic prompts
Use cases of Workflows MCP
Creating a 'production incident workflow' to analyze and resolve incidents
Automating code debugging with a 'debugger mode' workflow
Streamlining GitHub tasks with 'PR review' and 'PR creation' workflows
Improving reasoning on any task with 'thinking mode' workflows
Systematic code change planning with codebase analysis
FAQ from Workflows MCP
How do I install Workflows MCP?
How do I install Workflows MCP?
Use the npx -y @agentdesk/workflows-mcp@latest command or the provided JSON configuration in your MCP client.
How do I configure custom workflows?
How do I configure custom workflows?
Create a .workflows or .mcp-workflows directory in your project and add YAML configuration files.
How do I use presets?
How do I use presets?
Use the --preset flag when starting the server to specify which presets to load (e.g., --preset thinking,coding).
How do I inject parameters into prompts?
How do I inject parameters into prompts?
Define parameters in your YAML configuration and use the {{ parameter_name }} syntax in your prompts.
What happens if I update my configuration?
What happens if I update my configuration?
You must refresh the MCP tool in your client. If refreshing doesn't work, ensure your YAML is valid or try removing and re-adding the tool.