Prompts and Resources

Understanding MCP Resources and Prompts - advanced features that help AI agents discover and learn.

What Are MCP Resources?

Resources are named, readable content that AI agents can discover and browse directly, similar to how humans browse documentation.

Key Differences from Tools

Feature	Tools	Resources
Purpose	Execute commands	Read documentation
Input	Required parameters	Resource URI
Output	Command results	Document content
Use Case	Actions	Learning and discovery
Agent Behavior	Call with parameters	Browse and read

How Agents Use Resources

Instead of searching or guessing:

Agent: "I want to import data but I'm not sure how"

Traditional (without resources):
- Agent calls hana_search → finds import.md
- Agent guesses at parameters
- Agent tries import (may fail)

With Resources:
- Agent lists available resources
- Agent reads hana://docs/commands/import
- Agent learns correct usage
- Agent succeeds first time

Available Resources

Core Documentation Resources

Project Overview

• hana://docs/overview - Project introduction and features
• hana://docs/getting-started - Installation and setup guide

Connection and Setup

• hana://docs/connection-guide - 7-step connection resolution
• hana://docs/security - Security best practices
• hana://docs/parameters - Standard parameter conventions

Architecture and Design

• hana://docs/best-practices - Naming conventions and patterns
• hana://docs/project-structure - Project folder organization
• hana://docs/implementation - Technical implementation details

Category Guides

Organized by functionality:

• hana://docs/categories/data-quality - Data validation, profiling, duplicates
• hana://docs/categories/performance - Performance analysis and tuning
• hana://docs/categories/data-operations - Import, export, sync
• hana://docs/categories/inspection - Schema, table, view exploration
• hana://docs/categories/backup - Backup and recovery operations
• hana://docs/categories/security - User management and security
• hana://docs/categories/btp - SAP BTP integration

Command Documentation

Individual command guides (available for all 150+ commands):

• hana://docs/commands/import - Import command with examples
• hana://docs/commands/export - Export command with examples
• hana://docs/commands/dataValidator - Data validation guide
• hana://docs/commands/dataProfile - Data profiling guide
• hana://docs/commands/compareSchema - Schema comparison guide
• hana://docs/commands/[any-command] - Any command documentation

Each command resource includes:

• Detailed description
• All parameters explained
• Use cases and examples
• Common issues and solutions
• Related commands

Examples and Presets

• hana://examples/import - 5 real-world import scenarios
• hana://examples/export - 3 export scenarios
• hana://examples/data-migration - Migration examples
• hana://presets/safe-import - Safe import parameter template
• hana://presets/quick-export - Quick export template

Workflow Resources

Pre-built task sequences:

• hana://workflows/data-quality-check - Data quality check workflow
• hana://workflows/schema-migration - Schema migration workflow
• hana://workflows/performance-baseline - Performance baseline workflow
• hana://prompts - All available prompts and guide workflows

What Are MCP Prompts?

Prompts are guided conversation templates that help AI agents follow structured workflows for common tasks.

How Prompts Work

Prompts provide:

1. Multi-step guidance - Step-by-step instructions
2. Context preservation - Information carries through steps
3. Best practices - Baked-in standards
4. Error prevention - Validation and checks
5. Learning support - Explanations and tips

Example Prompt Flow

User: "Help me safely import customer data"

System: Invokes import-data prompt

Prompt Steps:

📋 SAFE DATA IMPORT WORKFLOW

Step 1: Verify Source File
└─ Review file location and format
   └─ Expected: CSV, Excel, or TSV file

Step 2: Inspect Target Table
└─ Examine table structure and constraints
   └─ Required: INSERT privilege on table

Step 3: Preview Import (Dry Run)
└─ Run import in preview mode
   └─ Shows what would be imported without actual changes

Step 4: Review Errors
└─ Check dry run results
   └─ Decide: Proceed or fix data issues

Step 5: Execute Import
└─ Run actual import with selected options
   └─ Progress tracking and error handling

Step 6: Validate Results
└─ Verify imported data
   └─ Check count and sample records
   └─ Run quality checks if needed

Available Prompts

1. Explore Database (explore-database)

Duration: 15-30 minutes

Parameters:

• schema (optional) - Specific schema to explore

Guided Steps:

1. Verify database connection
2. Check database version and system info
3. List all schemas
4. For each interesting schema:
   • List tables
   • Inspect some table structures
5. Profile data quality if interested
6. Summarize findings

Outcomes:

• Understanding of database structure
• Schema catalog
• Sample table definitions
• Data quality overview

2. Import Data Safely (import-data)

Duration: 20-40 minutes

Parameters:

• filename (required) - File to import
• table (optional) - Target table
• schema (optional) - Target schema

Guided Steps:

1. Verify file exists and is readable
2. Inspect target table structure
3. Preview import with dry-run
4. Review and resolve errors
5. Execute actual import (if approved)
6. Validate imported data
7. Generate import report

Outcomes:

• Successful safe import
• Error documentation
• Validation report

3. Troubleshoot Connection (troubleshoot-connection)

Duration: 20-40 minutes

Parameters: None required

Guided Steps:

1. Check basic connectivity
2. Verify credentials
3. Test database connection
4. Review user privileges
5. Check schema access
6. Diagnose specific issues
7. Provide remediation steps

Outcomes:

• Diagnosed connection issue
• Recommended solutions
• Verified working connection

4. Validate Data Quality (validate-data-quality)

Duration: 30-60 minutes

Parameters:

• table (required) - Table to validate
• schema (optional) - Target schema

Guided Steps:

1. Profile the table (data distribution, nulls, etc.)
2. Check for duplicate records
3. Run data validator
4. Analyze issues found
5. Get recommendations
6. Generate quality report

Outcomes:

• Data quality assessment
• Issue prioritization
• Remediation recommendations

5. Quick Start (quickstart)

Duration: 15-30 minutes

Parameters: None required

Perfect for: First-time users

Teaches:

1. hana_status - Verify connection
2. hana_version - Check database version
3. hana_schemas - List schemas
4. hana_tables - Explore tables
5. hana_inspectTable - View table structure
6. hana_healthCheck - System health

Outcomes:

• Understanding of basic commands
• Confidence in CLI usage
• Ready for advanced workflows

6. Export Data Safely (export-data)

Duration: 20-40 minutes

Parameters:

• table (required) - Table to export
• schema (optional) - Source schema
• format (optional) - CSV, Excel, or TSV

Guided Steps:

1. Verify source table exists
2. Check user has SELECT privilege
3. Configure export format
4. Preview export
5. Execute export
6. Verify export file
7. Validate data integrity

Outcomes:

• Successfully exported file
• Export validation report
• Format verification

How Agents Use Resources and Prompts

Resource Discovery Workflow

1. Agent: "I need help with..."
   
2. System: Lists available resources
   - hana://docs/getting-started
   - hana://docs/commands/import
   - hana://examples/import
   - hana://workflows/data-migration
   
3. Agent: "Show me hana://docs/commands/import"
   
4. System: Reads and formats resource
   - Title, description
   - All parameters explained
   - 5 usage scenarios
   - Common issues
   
5. Agent: Confident in tool usage
   
6. Agent: Runs import command

Prompt-Guided Workflow

1. User: "Help me explore the database"
   
2. Agent: Invokes explore-database prompt
   
3. System: Returns structured guidance
   Step 1: Verify connection
   Step 2: Check version
   Step 3: List schemas
   ...
   
4. Agent: Follows steps in order
   - Runs each command
   - Reviews results
   - Proceeds to next step
   
5. System: Automatically suggests next steps
   based on results
   
6. User: Gets clear understanding
   of database structure

Combining Resources and Prompts

Example: Learn Import by Resources, Execute by Prompt

Step 1: Resources (Learning)
- Agent reads hana://docs/commands/import
- Agent reviews hana://examples/import
- Agent understands options and best practices

Step 2: Prompt (Execution)
- Agent invokes "import-data" prompt
- Follows step-by-step guidance
- Ensures safe import

Step 3: Resources (Validation)
- If issues arise, agent reads hana://docs/troubleshooting
- Gets solutions from resources
- Applies fixes

Example: Performance Tuning

Step 1: Understand
- Agent reads hana://docs/categories/performance
- Agent reviews hana://workflows/performance-baseline
- Agent learns approach

Step 2: Execute
- Agent invokes relevant workflow
- Gets structured guidance
- Collects metrics

Step 3: Interpret
- Agent reads performance results
- Gets recommendations
- Plans improvements

Step 4: Implement & Verify
- Agent follows hana_get_template("performance-tuning")
- Implements suggestions
- Re-measures and compares

Best Practices

For Agent Developers

1. Start with Resource Discovery

   resources = await listResources()
   // Shows available learning material

2. Use Prompts for Complex Tasks

   await invokePrompt('import-data', { file: 'data.csv' })
   // Provides structured guidance

3. Chain Resources and Prompts

   Read resource → Invoke prompt → Execute → Validate

4. Leverage Context

   • Resources provide context
   • Prompts build on context
   • Results inform next steps

For Users (Best Practices)

1. Explore Resources First

   • Get familiar with available help
   • Learn best practices
   • See examples

2. Use Prompts for New Tasks

   • Follow guided workflows
   • Avoid mistakes
   • Learn as you work

3. Combine Both

   • Resources for learning
   • Prompts for execution
   • Both for confidence

Benefits

For AI Agents

✅ Reduced tool calls - Read documentation instead of searching ✅ Better context - Full information available ✅ Guided workflows - Structured multi-step processes ✅ Error prevention - Validation steps built in ✅ Better outcomes - Follows best practices

For Users

✅ Self-service learning - Agents can learn from resources ✅ Structured guidance - Prompts prevent mistakes ✅ Quick results - Faster task completion ✅ Best practices - Guided toward optimal approaches ✅ Confidence - Clear steps and expectations

Future Enhancements

Potential additions to resources and prompts:

1. Video Tutorials

   • hana://videos/import-guide
   • hana://videos/performance-tuning

2. Interactive Guides

   • Better step feedback
   • Real-time validation
   • Progress tracking

3. Custom Resources

   • User-created guides
   • Organization-specific docs
   • Custom workflows

4. Smart Recommendations

   • Context-aware prompts
   • Personalized workflows
   • Adaptive guidance

Next Steps

• Discovery Tools - Finding commands by intent
• Advanced Features - Workflows and interpretation
• Implementation Phases - Technical details
• Main Documentation Index - Overview of all features