An MCP (Model Context Protocol) server that provides LLMs with structured workflow guidance and "next step" instructions for development procedures. Perfect for maintaining consistent development practices across projects without having to repeatedly explain procedures to AI assistants.
This project welcomes contributions from AI/LLM agents! Pull requests from Claude, GPT, and other AI models are actively encouraged.
Guidelines for AI contributors:
- Follow existing code patterns and documentation standards
- Include comprehensive commit messages explaining changes
- Test changes thoroughly before submitting PRs
- Update documentation when adding new features
Engineering Manager MCP solves a common problem: LLMs forget your development procedures. Instead of repeatedly reminding your AI assistant about branch naming conventions, commit message formats, PR templates, or deployment steps, let Engineering Manager MCP provide that guidance on demand.
Use Cases:
- π Consistent Workflows - Maintain standardized development procedures across projects
- π Step-by-Step Guidance - LLMs can query "what's next?" at any point in the workflow
- π― Context-Aware Suggestions - Returns relevant workflow sections based on current task
- π§ Multi-Project Support - Different procedures for different projects
- π Template Variables - Customize procedures with project-specific values
# Run directly with uvx (recommended)
uvx engmanager-mcp
# Or install with pip
pip install engmanager-mcpFor project-specific configuration, add to your local .mcp.json in the project directory, or add to your global claude_code_config.json:
{
"mcpServers": {
"engmanager": {
"command": "uvx",
"args": ["engmanager-mcp"],
"env": {
"ENGMANAGER_DEFAULT_PROJECT": "myproject"
}
}
}
}- Create a project configuration:
mkdir -p ~/.config/engmanager-mcp
cat > ~/.config/engmanager-mcp/myproject-config.json <<EOF
{
"project_name": "myproject",
"procedure_file": "myproject-workflow.md",
"variables": {
"REPO_OWNER": "yourname",
"REPO_NAME": "myproject",
"DEFAULT_BRANCH": "main"
}
}
EOF- Create a workflow procedure:
cat > ~/.config/engmanager-mcp/myproject-workflow.md <<EOF
# My Project Workflow
## 1. Branch Creation
Create a feature branch:
\`\`\`bash
git checkout -b feature/my-feature
\`\`\`
## 2. Development
Make your changes and commit:
\`\`\`bash
git add .
git commit -m "feat: description"
\`\`\`
## 3. Push & PR
Push and create a pull request:
\`\`\`bash
git push -u origin feature/my-feature
gh pr create --title "feat: Description"
\`\`\`
EOF- Use in Claude:
You: "What's the first step in the workflow?"
Claude: [calls get_next_step tool]
Engineering Manager MCP: Returns "Step 1: Branch Creation..."
You: "What's next?"
Claude: [calls get_next_step with current_step=1]
Engineering Manager MCP: Returns "Step 2: Development..."
Get the next step in the workflow.
get_next_step(project="myproject", current_step=1)Parameters:
project(optional): Project name (uses default if not specified)current_step(optional): Current step number (defaults to 0 for first step)
Get a specific section from the workflow by name.
get_workflow_section(section="Error Recovery Protocols", project="myproject")Get an overview of all numbered steps.
list_workflow_steps(project="myproject")List all configured projects.
list_available_projects()Get detailed information about a project's configuration.
get_project_info(project="myproject")Get the full procedure file with variables substituted.
Get the project configuration.
Get documentation about template variables.
List all available projects with status.
Procedures support template variables for customization:
## Branch Creation for {PROJECT_NAME}
Create a branch:
\`\`\`bash
git checkout -b feature/my-feature
git push -u origin feature/my-feature
\`\`\`
Your repository: {REPO_OWNER}/{REPO_NAME}
Default branch: {DEFAULT_BRANCH}Common Variables:
{PROJECT_NAME}- Project name{REPO_OWNER}- GitHub repository owner{REPO_NAME}- GitHub repository name{DEFAULT_BRANCH}- Default branch (main/master){ISSUE_NUMBER}- Current issue number{BRANCH_PREFIX}- Branch prefix (feature/fix/refactor)
Define custom variables in your project's config file.
Important: The procedures/ directory is excluded from the PyPI package. When you install via uvx engmanager-mcp, you need to create your own workflow files in one of these locations:
~/.config/engmanager-mcp/ # Recommended for user-specific workflows
βββ myproject-config.json # Project configuration
βββ myproject-workflow.md # Workflow procedure
βββ another-project-config.json
βββ another-project-workflow.md
Configuration file locations (searched in order):
./procedures/(relative to current directory - for local development)~/.config/engmanager-mcp/(user config - recommended)/etc/engmanager-mcp/(system-wide)
Note: Example files (example-config.json, example-workflow.md) are only available in the GitHub repository for reference, not in the PyPI package.
ENGMANAGER_DEFAULT_PROJECT- Default project nameENGMANAGER_MCP_LOG_LEVEL- Logging level (DEBUG, INFO, WARNING, ERROR)ENGMANAGER_PROCEDURES_DIR- Custom procedures directoryENGMANAGER_MCP_TRANSPORT- Transport mode (stdio or http)
{
"project_name": "myproject",
"procedure_file": "myproject-workflow.md",
"variables": {
"REPO_OWNER": "username",
"REPO_NAME": "repository",
"DEFAULT_BRANCH": "main",
"CUSTOM_VAR": "custom_value"
}
}Required Fields:
project_name- Unique project identifierprocedure_file- Filename of the markdown procedure
Optional Fields:
variables- Dictionary of template variables
Problem: Different team members follow different git workflows.
Solution: Create a standardized procedure that LLMs can reference:
## 1. Branch Creation
- Always create from main
- Use conventional naming: feature/fix/refactor
- Format: <type>/<description>-issue-<number>
## 2. Commit Messages
- Use conventional commits format
- Include issue reference
- Add "Resolves #N" for auto-closeLLMs can now query these steps and follow them consistently.
Problem: Working on multiple projects with different procedures.
Solution: Configure multiple projects:
~/.config/engmanager-mcp/
βββ frontend-config.json
βββ frontend-workflow.md
βββ backend-config.json
βββ backend-workflow.mdLLMs can switch between project contexts easily.
Problem: New team members (including AI assistants) need workflow guidance.
Solution: Comprehensive procedure files serve as executable documentation:
## Developer Setup
1. Clone repository
2. Install dependencies
3. Configure environment
4. Run development server
## Development Workflow
[Steps...]
## Deployment Process
[Steps...]
## Error Recovery
[Procedures...]# Clone repository
git clone https://github.com/scarr7981/engmanager-mcp.git
cd engmanager-mcp
# Create virtual environment
python3 -m venv .venv
source .venv/bin/activate # Linux/Mac
# or
.venv\Scripts\activate # Windows
# Install dependencies
pip install -r requirements.txt
# Install in editable mode for development
pip install -e .
# Run locally
python -m engmanager_mcp.server
# or
engmanager-mcpFor local development and testing with Claude Code, add this to your project's .mcp.json file:
{
"mcpServers": {
"engmanager": {
"command": "/absolute/path/to/engmanager-mcp/.venv/bin/python",
"args": ["-m", "engmanager_mcp.server"],
"cwd": "/absolute/path/to/engmanager-mcp",
"env": {
"ENGMANAGER_DEFAULT_PROJECT": "example",
"ENGMANAGER_MCP_LOG_LEVEL": "DEBUG"
}
}
}
}Important:
- Replace
/absolute/path/to/engmanager-mcpwith the actual path to your cloned repository - For Windows WSL, use:
/mnt/c/Users/username/path/to/engmanager-mcp/.venv/bin/python - For Windows native, use:
C:\\absolute\\path\\to\\engmanager-mcp\\.venv\\Scripts\\python.exe - The
.mcp.jsonfile should be in the project root directory - This file is gitignored, so your local config won't be committed
This configuration:
- Uses the virtual environment's Python interpreter directly
- Runs from your local development directory
- Enables DEBUG logging for troubleshooting
- Sets the example project as default
- Allows you to edit code and see changes immediately (restart Claude Code to reload)
Create a test project:
mkdir -p procedures
cat > procedures/test-config.json <<EOF
{
"project_name": "test",
"procedure_file": "example-workflow.md",
"variables": {
"REPO_OWNER": "testuser",
"REPO_NAME": "testrepo",
"DEFAULT_BRANCH": "main"
}
}
EOF
# Copy example workflow
cp procedures/example-workflow.md procedures/test-workflow.md
# Test with MCP inspector or Claude Code# Install build tools
pip install build twine
# Build package
python -m build
# Check distribution
twine check dist/*# Upload to Test PyPI first
twine upload --repository testpypi dist/*
# Test installation
pip install --index-url https://test.pypi.org/simple/ engmanager-mcp
# Upload to PyPI
twine upload dist/*This project can be configured with GitHub Actions for automatic PyPI publishing on tagged releases. See cargoshipper-mcp for reference.
This project is inspired by the EXAMPLE_PROCEDURE.md workflow for Trowel.io and follows similar patterns.
Contribution Guidelines:
- Follow existing code patterns
- Add comprehensive docstrings
- Update README for new features
- Test with real procedures
- Include example configurations
MIT License - See LICENSE file for details
- CargoShipper MCP - Infrastructure automation MCP server
- Model Context Protocol - Official MCP documentation
- FastMCP - Python MCP framework
- Use numbered steps for sequential workflows
- Include code examples in bash blocks
- Document error recovery procedures
- Add quality gates at key checkpoints
- Use template variables for project-specific values
- One config per project - Keep projects isolated
- Shared procedures - Reference common workflows
- Default project - Set for most common use case
- Version control - Keep procedures in git
- Ask "what's next?" - Simple queries work best
- Provide context - Mention current step if known
- Query sections - Jump to specific workflow parts
- List steps - Get overview before starting
- Check config file location
- Verify filename format:
<project>-config.json - Check
ENGMANAGER_DEFAULT_PROJECTenvironment variable
- Verify
procedure_filepath in config - Check procedures directory
- Ensure markdown file exists
- Check procedure file for
{VARIABLE}syntax - Add missing variables to config
variablesobject - Variables must be UPPERCASE with underscores
- Add at least one
<project>-config.jsonfile - Check config search paths
- Verify JSON syntax in config files
- Issues: GitHub Issues
- Documentation: GitHub README
- MCP Help: Model Context Protocol Docs
Made with β€οΈ for better AI-assisted development workflows