anortham/sherpa
If you are the rightful owner of sherpa and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcphub.com.
Sherpa is a Model Context Protocol (MCP) server designed to guide AI agents through systematic development workflows using behavioral adoption and positive reinforcement techniques.
๐๏ธ Sherpa - AI Development Workflow Guide
MCP server that guides AI agents through systematic development workflows using behavioral adoption and positive reinforcement
Sherpa is a Model Context Protocol (MCP) server that transforms AI-assisted development by guiding agents through proven workflows with behavioral adoption techniques. Using positive reinforcement, progress tracking, and contextual celebrations, Sherpa helps AI agents develop systematic coding practices that lead to higher quality outcomes.
Why Sherpa?
When working with AI coding assistants, it's easy for them to:
- Skip writing tests and rush to implementation
- Guess at solutions without systematic investigation
- Fix bugs without understanding root causes
- Create code without following proven patterns
Sherpa changes this through behavioral adoption:
- ๐ฏ Positive Reinforcement: Celebrates progress with encouraging feedback
- ๐ Progress Tracking: Monitors workflow completion and streak building
- ๐ Achievement System: Recognizes milestones and systematic development
- ๐ก Success Stories: Shares real-world examples from companies like Netflix, GitHub, and Shopify
- ๐ Dynamic Celebrations: Contextual encouragement based on current progress
Studies show systematic workflows reduce bugs by 60%+ and increase developer confidence dramatically.
Quick Start
1. Install Dependencies
Requirements: This project requires Bun for optimal performance.
# Install Bun if you haven't already:
curl -fsSL https://bun.sh/install | bash
# Install dependencies:
bun install
2. Initialize Sherpa
Run the setup script to create your global Sherpa configuration:
bun run setup
This creates ~/.sherpa/
with default workflows and logging directory. The workflows are copied to your home directory so you can customize them freely.
3. Configure Claude Desktop
Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json
on Mac):
{
"mcpServers": {
"sherpa": {
"command": "bun",
"args": ["run", "/absolute/path/to/sherpa-server.ts"]
}
}
}
Important: Use the absolute path to sherpa-server.ts
in this repository.
4. Start Coding!
In Claude, the AI will now have access to two powerful tools:
guide
- Expert guidance for your next step in the workflowapproach
- Choose and switch between development approaches
Your workflows are now available across all your projects! ๐
๐ฏ Behavioral Adoption Features
Sherpa uses proven behavioral psychology to help AI agents develop systematic habits:
Progress Tracking & Celebrations
- Step Completion: Celebrates each workflow step with contextual encouragement
- Phase Advancement: Special recognition when completing workflow phases
- Milestone Achievements: Unlocks achievements like "First Workflow Mastery" and "Workflow Veteran"
- Streak Building: Tracks consecutive days of systematic development
Dynamic Encouragement System
- Context-Aware Messages: Different celebrations for first steps vs. major milestones
- Workflow-Specific Praise: TDD gets test-focused encouragement, Bug Hunt gets detective metaphors
- Success Inspiration: Occasionally shares relevant company success stories
- Tool Usage Reinforcement: Positive feedback for using systematic approaches
Example Behavioral Flow
// After completing a test
{
"celebration": "๐งช Excellent! First test written - you're building bulletproof code!",
"tool_encouragement": "๐ฏ Excellent workflow awareness! Checking progress keeps you oriented.",
"progress_encouragement": "๐ Great progress! 1 workflows completed, 5 steps taken.",
"success_inspiration": "TDD practitioners at major companies ship 67% fewer bugs!"
}
Personalized Tips
Based on usage patterns, Sherpa provides personalized suggestions:
- Try different workflow types for varied experience
- Build consistency for better habit formation
- Complete more steps per workflow for maximum benefit
Available Workflows
๐งช TDD (tdd.yaml)
Pure test-first development. Perfect for:
- Building parsers, extractors, transformers
- Creating well-defined utilities
- Any feature where you know the expected behavior
๐ Bug Hunt (bug-hunt.yaml)
Systematic debugging with tests. Ideal for:
- Fixing reported bugs
- Investigating crashes
- Resolving test failures
๐ General (general.yaml)
Balanced approach with research โ plan โ test โ implement. Good for:
- Most feature development
- Adding new functionality
- Default workflow
๐ Rapid (rapid.yaml)
Quick prototyping with retroactive tests. Use for:
- Spikes and experiments
- Proof of concepts
- Demo preparation
๐ Planning (planning.yaml)
Pure planning workflow - research, understand, design, and document before implementation. Ideal for:
- Architecture design
- Feature specification
- Requirements analysis
- Research and documentation
โป๏ธ Refactor (refactor.yaml)
Safe refactoring with test coverage. Perfect for:
- Code cleanup
- Performance improvements
- Restructuring
Managing Workflows
Available Commands
# Check Sherpa status and installed workflows
bun run status
# Reinstall default workflows (preserves existing customizations)
bun run setup
# Force reinstall (overwrites customizations)
bun run setup:force
# Reset to default workflows (removes all customizations)
bun run reset
# View real-time logs
bun run logs
# View latest log file
bun run logs:latest
Customizing Workflows
Edit any file in ~/.sherpa/workflows/
to match your preferences:
name: "Your Workflow Name"
description: "What this workflow is for"
trigger_hints: # Keywords that auto-select this workflow
- "keyword1"
- "keyword2"
phases:
- name: "๐ Phase Name"
guidance: "What to focus on"
suggestions:
- "Specific action 1"
- "Specific action 2"
Creating New Workflows
- Create a new file:
~/.sherpa/workflows/my-workflow.yaml
- Follow the structure above
- Restart Claude to load the new workflow
Example Workflows
Check out workflows/examples/
for additional specialized workflows:
- ๐ Documentation - For writing guides and API docs
- ๐ Security Audit - For defensive security reviews
- ๐ Performance - For optimization projects
- ๐๏ธ Code Review - For thorough code reviews
Copy any example to your workflows directory:
cp ~/.sherpa/workflows/examples/documentation.yaml ~/.sherpa/workflows/
Usage Example
When Claude starts working on a task, Sherpa provides rich, encouraging guidance:
Claude: I need to fix the login bug.
[Calls: next check]
Sherpa: {
"workflow": "Bug Hunt",
"phase": "๐ Reproduce & Isolate",
"guidance": "Understand the bug completely before fixing",
"suggestions": ["Reproduce manually first", "Document exact steps", ...],
"tool_encouragement": "๐ฏ Excellent workflow awareness! Checking progress keeps you oriented.",
"progress_encouragement": "๐ฅ Amazing 3-day streak! You're building excellent workflow habits."
}
Claude: I'll systematically reproduce this issue first...
[After reproducing]
[Calls: next done: "Reproduced - happens with special characters in email"]
Sherpa: {
"celebration": "๐ Detective mode activated! Great systematic reproduction work.",
"workflow": "Bug Hunt",
"phase": "๐ฏ Capture in Test",
"guidance": "Lock down the bug with a failing test",
"suggestions": ["Write failing test with special chars", ...],
"progress": {"completed": 1, "total": 3, "remaining": 2},
"success_inspiration": "Netflix reduced critical incidents by 73% through systematic bug reproduction!"
}
Benefits in Action:
- ๐ฏ Claude naturally adopts systematic approaches
- ๐ Progress tracking builds momentum and confidence
- ๐ Celebrations reinforce good development habits
- ๐ก Success stories provide additional motivation
Tips for Success
- Start Simple: Use the provided workflows as-is initially
- Customize Gradually: Adjust workflows based on what works for your project
- Be Specific: Add clear, actionable suggestions to guide the AI
- Reinforce Good Habits: The AI will learn to use
guide
regularly
Philosophy
Sherpa transforms AI development through behavioral science:
Core Principles
- Positive Reinforcement: Celebrates systematic approaches rather than punishing shortcuts
- Progressive Habit Building: Gradually builds workflow discipline through consistent encouragement
- Context-Aware Guidance: Adapts feedback based on current progress and workflow type
- Evidence-Based: Incorporates real-world success stories from industry leaders
Design Philosophy
- Suggestive, not prescriptive: Guides without constraining creativity
- Lightweight: Just 2 intuitive tools, minimal context usage, maximum behavioral impact
- Customizable: Your workflows, your celebrations, your development culture
- Test-focused: Because AI-generated code needs systematic validation
- Joy-Driven: Makes systematic development feel rewarding and satisfying
Troubleshooting
Workflows Not Loading
- Ensure
~/.sherpa/workflows/
exists and contains YAML files - Run
bun run status
to check your installation - Check the logs:
bun run logs:latest
- If missing, run
bun run setup
to reinstall
AI Not Following Workflow
- Remind Claude to use the
guide
tool regularly - Sherpa will celebrate this! - The behavioral adoption system provides positive reinforcement automatically
- Consider adjusting workflow suggestions to be more specific and encouraging
- Check if celebrations and progress tracking are motivating consistent tool usage
Debugging Issues
- View real-time logs:
bun run logs
- Check server startup messages in Claude Desktop
- Verify the absolute path in your Claude Desktop config
- Logs are stored in
~/.sherpa/logs/
with automatic rotation (7 days)
Common Problems
"Sherpa not initialized" error:
bun run setup
"No workflows found" warning:
bun run reset # Reinstall default workflows
Server won't start:
- Check the absolute path in Claude Desktop config
- Ensure Bun is installed and in your PATH
- Check for TypeScript errors:
bun run sherpa-server.ts
Architecture
Behavioral Adoption System
Sherpa's behavioral adoption system consists of:
- Adaptive Learning Engine (
src/behavioral-adoption/adaptive-learning-engine.ts
): Tracks user patterns and provides personalized guidance - Progress Tracker (
src/behavioral-adoption/progress-tracker.ts
): Monitors workflow completion, tracks milestones, and maintains usage statistics - Celebration Generator (
src/behavioral-adoption/celebration-generator.ts
): Creates contextual encouragement and success celebrations - Encouragements System (
src/server-instructions/templates/encouragements.json
): 100+ contextual encouragement messages organized by workflow type and context
Core Architecture
- MCP Server (
sherpa-server.ts
): Main Model Context Protocol server with integrated behavioral system - Workflow Engine: YAML-based workflow definitions with phase-by-phase guidance
- File-based Logging: All behavioral events and progress logged to
~/.sherpa/logs/
Contributing
Ideas for improvements? Workflow patterns that work well? New behavioral adoption techniques? Please share!
Development
# Install dependencies
bun install
# Run tests
bun test
# Check server startup (for testing)
bun run start
# Initialize user workflows
bun run setup
# View logs during development
bun run logs
Remember: Like a good sherpa, this tool guides the journey but doesn't carry you. The AI still needs to do the work, but now it has a trusted, encouraging guide that celebrates every step forward.