Skip to main content

MCP Servers (Model Context Protocol)

Model Context Protocol (MCP) servers extend Chatty AI with additional capabilities like filesystem access, Git operations, and database connections.

Overview​

MCP is an open protocol that enables AI assistants to securely access external data sources and tools. Chatty AI supports connecting to multiple MCP servers simultaneously.

Common Use Cases:

  • File system operations (read, write, search)
  • Git repository management
  • Database queries
  • API integrations
  • Custom tool development

Available MCP Servers​

Chatty AI Provided MCP Servers​

1. Microsoft 365 MCP Server​

Chatty AI multitenant app integration - No Azure setup required!

Capabilities:

  • Outlook email access
  • OneDrive file operations
  • Calendar management
  • Teams integration
  • SharePoint access
  • To Do tasks
  • OneNote notebooks
  • Excel operations

MCP Server URL:

# Chatty AI Hosted (Recommended)
https://mcp-ms365.chattyai.io

# Or internal deployment
http://mcp-ms365:3100

Configuration:

# In Chatty AI Admin UI
Settings → MCP Servers → Add Server
Name: Microsoft 365
URL: https://mcp-ms365.chattyai.io
Description: MS365 integration via Chatty AI multitenant app

Benefits:

  • ✅ No Azure App Registration needed
  • ✅ Pre-configured permissions
  • ✅ Vendor-managed credentials
  • ✅ All MS365 services enabled

See Microsoft 365 Apps Integration for details.

2. Jira MCP Server​

Jira integration via Chatty AI MCP server

Capabilities:

  • Create and update issues
  • Query projects
  • Search issues
  • Add comments
  • Transition workflows

MCP Server URL:

# Chatty AI Hosted
https://mcp-jira.chattyai.io

# Or internal deployment
http://mcp-jira:3101

Configuration:

# In Chatty AI Admin UI
Settings → MCP Servers → Add Server
Name: Jira
URL: https://mcp-jira.chattyai.io

See Jira Integration for details.

3. Monday.com MCP Server​

Monday.com integration via Chatty AI MCP server

Capabilities:

  • Create and update items
  • Query boards
  • Manage columns
  • Add updates
  • Track progress

MCP Server URL:

# Chatty AI Hosted
https://mcp-monday.chattyai.io

# Or internal deployment
http://mcp-monday:3102

Configuration:

# In Chatty AI Admin UI
Settings → MCP Servers → Add Server
Name: Monday.com
URL: https://mcp-monday.chattyai.io

See Monday.com Integration for details.

General Purpose MCP Servers​

4. Filesystem MCP Server​

Access and manipulate files on the server.

Capabilities:

  • Read file contents
  • Write files
  • List directories
  • Search files
  • File metadata

MCP Server URL:

http://mcp-filesystem:3000

Configuration:

# In Chatty AI Admin UI
Settings → MCP Servers → Add Server
Name: Filesystem
URL: http://mcp-filesystem:3000

5. Git MCP Server​

Interact with Git repositories.

Capabilities:

  • Clone repositories
  • Read file contents from repos
  • View commit history
  • Branch operations
  • Diff viewing

MCP Server URL:

http://mcp-git:3001

Configuration:

# In Chatty AI Admin UI
Settings → MCP Servers → Add Server
Name: Git
URL: http://mcp-git:3001

6. Database MCP Server​

Query databases directly.

Capabilities:

  • Execute SQL queries
  • List tables and schemas
  • View table data
  • Database metadata

MCP Server URL:

http://mcp-database:3002

Configuration:

# In Chatty AI Admin UI
Settings → MCP Servers → Add Server
Name: Database
URL: http://mcp-database:3002

7. Custom MCP Servers​

Deploy your own MCP servers for custom functionality.

Example Custom Server:

http://your-custom-mcp:3003

Configuration Methods​

  1. Login to Chatty AI as admin
  2. Navigate to Settings → MCP Servers
  3. Click "Add Server"
  4. Fill in details:
    • Name: Descriptive name (e.g., "Filesystem")
    • URL: MCP server URL (e.g., http://mcp-filesystem:3000)
    • Description: Optional description
  5. Click "Save"
  6. Test connection

Method 2: Environment Variables​

Configure default MCP servers via environment variables:

# Add to .env file
MCP_SERVERS='[
  {
    "name": "Filesystem",
    "url": "http://mcp-filesystem:3000",
    "enabled": true
  },
  {
    "name": "Git",
    "url": "http://mcp-git:3001",
    "enabled": true
  },
  {
    "name": "Database",
    "url": "http://mcp-database:3002",
    "enabled": true
  }
]'

Restart services:

docker compose restart chattyai

Method 3: API Configuration​

Use Chatty AI API to configure MCP servers:

# Add MCP server via API
curl -X POST https://chat.example.com/api/mcp/servers \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Filesystem",
    "url": "http://mcp-filesystem:3000",
    "enabled": true
  }'

Deploying MCP Servers​

Using Docker Compose​

Add MCP servers to your docker-compose.yaml:

services:
  # Filesystem MCP Server
  mcp-filesystem:
    image: modelcontextprotocol/filesystem-server:latest
    container_name: mcp-filesystem
    ports:
      - "3000:3000"
    volumes:
      - /path/to/files:/data:ro  # Read-only access
    environment:
      - MCP_PORT=3000
      - MCP_ROOT_PATH=/data
    networks:
      - chatty-network

  # Git MCP Server
  mcp-git:
    image: modelcontextprotocol/git-server:latest
    container_name: mcp-git
    ports:
      - "3001:3001"
    environment:
      - MCP_PORT=3001
    networks:
      - chatty-network

  # Database MCP Server
  mcp-database:
    image: modelcontextprotocol/database-server:latest
    container_name: mcp-database
    ports:
      - "3002:3002"
    environment:
      - MCP_PORT=3002
      - DB_HOST=your-db-host
      - DB_PORT=5432
      - DB_USER=readonly_user
      - DB_PASSWORD=readonly_password
    networks:
      - chatty-network

networks:
  chatty-network:
    external: true

Deploy:

docker compose up -d mcp-filesystem mcp-git mcp-database

Security Considerations​

1. Network Isolation​

MCP servers should be on the same Docker network as Chatty AI:

networks:
  chatty-network:
    external: true

2. Read-Only Access​

For filesystem access, use read-only volumes:

volumes:
  - /path/to/files:/data:ro  # :ro = read-only

3. Dedicated Service Accounts​

For database MCP servers, use read-only database users:

-- Create read-only user
CREATE USER mcp_readonly WITH PASSWORD 'secure_password';
GRANT CONNECT ON DATABASE mydb TO mcp_readonly;
GRANT SELECT ON ALL TABLES IN SCHEMA public TO mcp_readonly;

4. Authentication​

Secure MCP servers with authentication:

# MCP server with API key
MCP_API_KEY=your-secure-api-key

# Configure in Chatty AI
Settings → MCP Servers → Add Server
URL: http://mcp-server:3000
Headers: {"Authorization": "Bearer your-secure-api-key"}

5. HTTPS/TLS​

For production, use HTTPS for MCP server communication:

# MCP server URL with HTTPS
https://mcp-server.example.com

Using MCP Servers in Chatty AI​

Example: Reading Files​

User: Can you read the contents of /data/config.yaml?

Chatty AI: [Uses Filesystem MCP Server]
Here are the contents of config.yaml:
...

Example: Git Operations​

User: Show me the latest commits in the main branch

Chatty AI: [Uses Git MCP Server]
Latest commits:
1. Fix bug in authentication (2 hours ago)
2. Add new feature (1 day ago)
...

Example: Database Queries​

User: How many users are in the database?

Chatty AI: [Uses Database MCP Server]
There are 1,234 users in the database.

Troubleshooting​

MCP Server Not Connecting​

# Check MCP server is running
docker ps | grep mcp

# Check MCP server logs
docker logs mcp-filesystem

# Test MCP server directly
curl http://mcp-filesystem:3000/health

# Check network connectivity
docker exec chattyai ping mcp-filesystem

Permission Denied​

# Check volume permissions
ls -la /path/to/files

# Fix permissions
sudo chown -R 1000:1000 /path/to/files

# Or use read-only access
volumes:
  - /path/to/files:/data:ro

MCP Server Timeout​

# Increase timeout in Chatty AI
Settings → MCP Servers → Edit Server
Timeout: 30000  # 30 seconds

# Or in environment variables
MCP_TIMEOUT=30000

Custom MCP Server Development​

Creating a Custom MCP Server​

Example Node.js MCP server:

// server.js
const express = require('express');
const app = express();

app.get('/health', (req, res) => {
  res.json({ status: 'ok' });
});

app.post('/tools/my-tool', (req, res) => {
  // Your custom tool logic
  res.json({ result: 'success' });
});

app.listen(3003, () => {
  console.log('Custom MCP server running on port 3003');
});

Dockerfile:

FROM node:18-alpine
WORKDIR /app
COPY package*.json ./
RUN npm install
COPY . .
EXPOSE 3003
CMD ["node", "server.js"]

Deploy:

docker build -t my-custom-mcp .
docker run -d --name my-custom-mcp \
  --network chatty-network \
  -p 3003:3003 \
  my-custom-mcp

MCP Server Checklist​

Setup​

  • MCP servers deployed
  • Network connectivity verified
  • Servers added in Chatty AI admin UI
  • Connection tested
  • Permissions configured

Security​

  • Read-only access where appropriate
  • Authentication configured
  • Network isolation implemented
  • HTTPS enabled (production)
  • Service accounts created

Testing​

  • Health check passing
  • Tools accessible from Chatty AI
  • Permissions working correctly
  • Error handling tested
  • Performance acceptable