Sean-m-e/jess-mcp-server
If you are the rightful owner of jess-mcp-server and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to dayong@mcphub.com.
The Jess MCP Server v1.0.0 is a Model Context Protocol server designed for the Jess voice agent v3.33, providing an externalized knowledge base and behavioral protocols for OBrien Real Estate Brighton.
Jess MCP Server v1.0.0
Model Context Protocol (MCP) server for Jess voice agent v3.33, providing externalized knowledge base and behavioral protocols for OBrien Real Estate Brighton.
Overview
This MCP server externalizes 8 major sections from the Jess v3.33 prompt (1,024 lines extracted), reducing the main prompt size by providing on-demand access to:
- Knowledge Base (96 lines) - Q&A responses for customer queries
- Pronunciation Rules (138 lines) - Phonetic guidelines for Australian locations and monetary values
- Speech Enhancement (45 lines) - Natural response variations and contextual matching
- Australian Patterns (82 lines) - Recognition patterns for Australian English
- Behavioral Protocols (237 lines) - Phase control and continuous information extraction
- Contextual Understanding (56 lines) - Complex response handling
- Closing Protocols (67 lines) - Call closing logic for all outcomes
- Reactive Modules (303 lines) - Modules R1-R24 for specific scenarios
Architecture
jess-mcp-server/
├── src/
│ ├── index.ts # Main MCP server
│ └── data/ # Extracted JSON data files
│ ├── knowledge.json
│ ├── pronunciation.json
│ ├── speech_enhancement.json
│ ├── australian_patterns.json
│ ├── behavioral_protocols.json
│ ├── contextual_understanding.json
│ ├── closing_protocols.json
│ └── reactive_modules.json
├── package.json
├── tsconfig.json
├── railway.json # Railway deployment config
└── README.md
Quick Start
Local Development
-
Install dependencies:
npm install -
Build the server:
npm run build -
Run the server:
npm start -
Development mode (with watch):
npm run watch
Railway Deployment
Prerequisites
- GitHub account
- Railway account (https://railway.app)
- Git installed locally
Deployment Steps
-
Initialize Git repository:
git init git add . git commit -m "Initial MCP server for Jess v3.33" -
Push to GitHub:
# Create new repository on GitHub first, then: git remote add origin https://github.com/YOUR_USERNAME/jess-mcp-server.git git branch -M main git push -u origin main -
Deploy on Railway:
- Go to https://railway.app/dashboard
- Click "New Project"
- Select "Deploy from GitHub repo"
- Choose your
jess-mcp-serverrepository - Railway will auto-detect Node.js and deploy
- Wait for deployment to complete
-
Get your MCP server URL:
- Once deployed, Railway will provide a URL
- Copy this URL for voice platform integration
-
Configure in Voice Platform:
- Use Railway deployment URL as "MCP ID"
- Add to voice platform MCP settings
- Test connection
- Verify resources are accessible
Available Resources
The server exposes 8 resources via MCP protocol:
| URI | Name | Description |
|---|---|---|
jess://knowledge/all | Knowledge Base | Q&A responses for customer queries |
jess://linguistic/pronunciation_rules | Pronunciation Rules | Phonetic guidelines for locations and values |
jess://linguistic/speech_enhancement | Speech Enhancement | Natural response variations |
jess://linguistic/australian_patterns | Australian Patterns | Australian English recognition |
jess://protocols/global_behavioral | Behavioral Protocols | Phase control and extraction rules |
jess://protocols/contextual_understanding | Contextual Understanding | Complex response handling |
jess://protocols/unified_closing | Closing Protocols | Call closing logic |
jess://modules/reactive/all | Reactive Modules | Modules R1-R24 |
Testing
Test Resource Availability
After deployment, you can test resource availability:
# List all resources
curl -X POST https://your-railway-url/mcp \
-H "Content-Type: application/json" \
-d '{"method": "resources/list"}'
# Read specific resource
curl -X POST https://your-railway-url/mcp \
-H "Content-Type: application/json" \
-d '{"method": "resources/read", "params": {"uri": "jess://knowledge/all"}}'
Integration with Voice Platform
- Copy Railway URL from Railway dashboard
- Add to voice platform MCP settings as "MCP ID"
- Configure resources that Jess should access
- Test connection to verify accessibility
Content Integrity
All extracted content is byte-for-byte identical to source (JESS_V3_33_FIXED_OPTION_A.txt):
- Total lines extracted: 1,024 lines
- Total characters: 43,998 characters
- JSON validation: All 8 files validated successfully
- Escaping verified: Special characters properly escaped
Validation Reports
See /mnt/user-data/outputs/validation/ for:
EXTRACTION_LOG.json- Detailed extraction logSECTION_BOUNDARIES.json- Exact line numbers for each sectionCHARACTER_COUNT_LOG.json- Character count verificationJSON_INTEGRITY_REPORT.json- JSON validation results
Version Compatibility
- Jess Prompt Version: v3.33+
- MCP Server Version: 1.0.0
- Node.js: >= 18.0.0
- MCP SDK: ^0.5.0
Rollback Strategy
If issues occur in production:
- Disable MCP integration in voice platform
- Revert to Jess v3.33 (full inline prompt)
- Investigate issue using validation logs
- Fix and redeploy when ready
Troubleshooting
Build fails
- Check Node.js version (>= 18.0.0)
- Verify all dependencies installed:
npm install - Check TypeScript compilation:
npm run build
Railway deployment fails
- Verify
railway.jsonis present - Check build logs on Railway dashboard
- Ensure
package.jsonscripts are correct
Resources not accessible
- Verify server is running: Check Railway logs
- Check resource URIs match expected format
- Test with curl commands above
Content corruption
- Check JSON files in
src/data/directory - Verify character counts match extraction logs
- Re-run validation scripts
Support
For issues or questions:
- Check validation logs in
/mnt/user-data/outputs/validation/ - Review Railway deployment logs
- Verify MCP integration in voice platform
License
MIT License - See project documentation for details.
Version: 1.0.0
Date: 2025-11-17
Source: Jess v3.33 (JESS_V3_33_FIXED_OPTION_A.txt)
Total Extracted: 1,024 lines across 8 sections