Spec-Driven IntelliMatch / iSuite MCP
Enables interaction with IntelliMatch recruiting and iSuite insurance systems through spec-driven MCP tools, allowing natural language queries for candidate search, policy management, claims, and workflows.
README
Spec-Driven IntelliMatch / iSuite MCP
A single Model Context Protocol server for both IntelliMatch and iSuite, where every tool is defined in a YAML or JSON file — no code changes needed to add new capabilities.
Architecture
specs/
├── intellimatch/ ← IntelliMatch tool specs
│ ├── candidates.yaml
│ ├── jobs.yaml
│ └── reports.yaml
└── isuite/ ← iSuite tool specs
├── policies.yaml
├── claims.yaml
└── workflows.yaml
config/
└── connections.yaml ← API/DB/log connection definitions
src/
├── loader/ ← Scans and validates spec files
├── connections/ ← Connection pool manager
├── executor/ ← Action handlers (http, db, log, ui)
├── template.ts ← {{variable}} interpolation engine
└── server.ts ← MCP server (list/call tools)
Adding a new tool = adding or editing a YAML file. Zero code changes.
Quick Start
1. Install dependencies
npm install
For UI automation (optional):
npx playwright install chromium
2. Configure environment
cp .env.example .env
# Edit .env with your actual URLs, credentials, and DB connection strings
3. Configure connections
Edit config/connections.yaml to point at your IntelliMatch and iSuite environments. Connection values support ${ENV_VAR} syntax — keep secrets in .env, not the YAML.
4. Build and run
npm run build # Compile TypeScript → dist/
npm start # Run the MCP server (stdio transport)
Development (no build step):
npm run dev # Run directly with tsx
Inspect tools interactively:
npm run inspect # Opens MCP Inspector in browser
VS Code Integration
Option A — Claude Code (recommended)
Add the server to your Claude Code MCP config. In VS Code, open the Command Palette → Claude: Open MCP Settings, or edit the file directly:
.claude/settings.json (project-level, checked in):
{
"mcpServers": {
"intellimatch-isuite": {
"command": "node",
"args": ["C:/NewInitiatives/mcp/Spec-Driven Intellimatch MCP/dist/index.js"],
"env": {
"INTELLIMATCH_API_URL": "https://intellimatch.internal/api",
"INTELLIMATCH_API_TOKEN": "your-token",
"ISUITE_API_URL": "https://isuite.internal/api",
"ISUITE_USERNAME": "api_user",
"ISUITE_PASSWORD": "your-password",
"INTELLIMATCH_DB_HOST": "localhost",
"INTELLIMATCH_DB_NAME": "intellimatch",
"INTELLIMATCH_DB_USER": "im_readonly",
"INTELLIMATCH_DB_PASSWORD": "your-password",
"ISUITE_DB_HOST": "sqlserver.internal",
"ISUITE_DB_NAME": "iSuiteDB",
"ISUITE_DB_USER": "isuite_readonly",
"ISUITE_DB_PASSWORD": "your-password"
}
}
}
}
Or use tsx for development (no build step):
{
"mcpServers": {
"intellimatch-isuite": {
"command": "npx",
"args": [
"tsx",
"C:/NewInitiatives/mcp/Spec-Driven Intellimatch MCP/src/index.ts"
]
}
}
}
Option B — VS Code MCP Extension
If you're using the VS Code MCP extension, add to settings.json:
{
"mcp.servers": {
"intellimatch-isuite": {
"command": "node",
"args": ["${workspaceFolder}/dist/index.js"],
"transport": "stdio"
}
}
}
Adding New Tools
- Create or edit a YAML file in
specs/intellimatch/orspecs/isuite/ - Add a tool entry following the spec format below
- Restart the MCP server — it reloads all specs on startup
Spec File Format
app: intellimatch # or isuite, or shared
version: "1.0"
tools:
- name: im_my_tool # lowercase_with_underscores, prefix im_ or is_
description: >
What this tool does — shown to Claude so write it clearly.
inputSchema:
type: object
properties:
my_param:
type: string
description: What this param does
limit:
type: integer
default: 20
required:
- my_param
action:
type: http | db | log | ui # pick one
# ... action-specific config below
Action Types
http — REST API call
action:
type: http
connection: intellimatch_api # name from config/connections.yaml
method: GET # GET POST PUT PATCH DELETE
path: /v1/candidates/{{candidate_id}}
params: # URL query params (GET)
q: "{{query}}"
limit: "{{limit}}"
body: # Request body (POST/PUT)
name: "{{name}}"
headers: # Extra headers (optional)
X-Custom: "{{custom_value}}"
resultPath: data.items # dot-path to extract from response (optional)
{{variable}} patterns are resolved from the tool's input arguments.
db — Parameterized SQL query
action:
type: db
connection: intellimatch_db # postgres | mssql | mysql connection
# Write native SQL for your DB type. NEVER put {{variables}} in the query.
query: >
SELECT id, name, status FROM candidates
WHERE status = $1 AND name ILIKE '%' || $2 || '%'
LIMIT $3
params:
- "{{status}}" # resolved and passed as $1 (pg) / ? (mysql) / @param1 (mssql)
- "{{query}}" # → $2
- "{{limit}}" # → $3
rowLimit: 500 # hard cap on returned rows
SQL injection safe: user inputs in params are passed as parameterized values, never string-interpolated into the SQL.
PostgreSQL: use $1, $2, ...
MySQL: use ?, ?, ...
SQL Server: use @param1, @param2, ...
log — Log access
File-based (path must be under LOG_BASE_DIR):
action:
type: log
source: file
path: "/var/log/intellimatch/app.log"
lines: 200
filter: "{{search_term}}" # optional grep-style filter
HTTP log API:
action:
type: log
source: http
connection: intellimatch_logs
endpoint: /logs/search
params:
level: "{{level}}"
limit: "{{lines}}"
ui — Browser automation (Playwright)
action:
type: ui
connection: intellimatch_ui # type: ui connection with baseUrl and auth
browser: chromium # chromium | firefox | webkit
headless: true
steps:
- action: navigate
url: "{{connection.baseUrl}}/login"
- action: fill
selector: "#username"
value: "{{connection.auth.username}}"
- action: fill
selector: "#password"
value: "{{connection.auth.password}}"
- action: click
selector: "[type=submit]"
- action: waitFor
selector: ".dashboard"
timeout: 10000
- action: navigate
url: "{{connection.baseUrl}}/reports"
- action: screenshot
fullPage: false
- action: getText
selector: ".result-count"
as: resultCount # stored in output as results.resultCount
- action: getTable
selector: "table.data"
as: tableData
- action: selectOption
selector: "#status-filter"
value: "{{status}}"
{{connection.baseUrl}}, {{connection.auth.username}}, and {{connection.auth.password}} are automatically resolved from the connection config.
Connection Configuration
Edit config/connections.yaml. All ${ENV_VAR} references are resolved from environment variables at startup.
HTTP Connection
my_api:
type: http
baseUrl: "${MY_API_URL}"
timeout: 30000
headers:
Accept: application/json
auth:
type: bearer # bearer | basic | api-key | none
token: "${MY_TOKEN}"
PostgreSQL
my_pg_db:
type: postgres
host: "${PG_HOST}"
port: 5432
database: "${PG_DB}"
user: "${PG_USER}"
password: "${PG_PASSWORD}"
ssl: true
pool:
min: 1
max: 5
SQL Server
my_mssql:
type: mssql
server: "${MSSQL_HOST}"
database: "${MSSQL_DB}"
user: "${MSSQL_USER}"
password: "${MSSQL_PASSWORD}"
options:
encrypt: true
trustServerCertificate: false
MySQL / MariaDB
my_mysql:
type: mysql
host: "${MYSQL_HOST}"
port: 3306
database: "${MYSQL_DB}"
user: "${MYSQL_USER}"
password: "${MYSQL_PASSWORD}"
UI (Playwright)
my_app_ui:
type: ui
baseUrl: "${APP_UI_URL}"
auth:
username: "${UI_USERNAME}"
password: "${UI_PASSWORD}"
Useful Claude Prompts
Once the MCP is connected in Claude Code, try these prompts:
IntelliMatch — Recruiting
Search for active Java developer candidates in London and show the top 10 matches.
Find the best candidates for job requisition JOB-2025-0042 with a match score above 0.7.
What is the current pipeline status for candidate C-10023?
Show me the placement metrics for Q1 2025 broken down by department.
Get the last 50 error logs from the IntelliMatch matching engine.
Take a screenshot of the IntelliMatch reports dashboard.
Which jobs have been open for more than 30 days and are still unfilled?
iSuite — Insurance
Search for all active home insurance policies for "John Smith".
Get the full details for policy POL-2025-001234 including coverage and renewal date.
Show me all open claims filed this year with their current amounts and status.
Submit a new auto insurance claim for policy POL-2025-001234 — incident on 2025-06-10, minor collision, estimated $3,500.
List all underwriting workflows pending my review.
Approve workflow WF-2025-00891 with the note "All requirements verified, approved for standard rate".
Show the claims summary for 2025 — total count, average settlement, and days to close by claim type.
What iSuite workflows have been stalled for more than 5 days?
Take a screenshot of workflow WF-2025-00891 for the approval documentation.
General
List all available IntelliMatch and iSuite tools and what they do.
What tools are available for working with insurance claims?
Environment Variables Reference
| Variable | Required | Description |
|---|---|---|
INTELLIMATCH_API_URL |
Yes | IntelliMatch REST API base URL |
INTELLIMATCH_API_TOKEN |
Yes | Bearer token for IntelliMatch API |
INTELLIMATCH_DB_HOST |
If using DB tools | PostgreSQL host |
INTELLIMATCH_DB_NAME |
If using DB tools | Database name |
INTELLIMATCH_DB_USER |
If using DB tools | DB username |
INTELLIMATCH_DB_PASSWORD |
If using DB tools | DB password |
ISUITE_API_URL |
Yes | iSuite REST API base URL |
ISUITE_USERNAME |
Yes | iSuite API username |
ISUITE_PASSWORD |
Yes | iSuite API password |
ISUITE_DB_HOST |
If using DB tools | SQL Server host |
ISUITE_DB_NAME |
If using DB tools | Database name |
ISUITE_DB_USER |
If using DB tools | DB username |
ISUITE_DB_PASSWORD |
If using DB tools | DB password |
INTELLIMATCH_LOGS_URL |
If using log tools | Log API base URL |
INTELLIMATCH_LOGS_TOKEN |
If using log tools | Log API token |
LOG_BASE_DIR |
If using file logs | Root directory for log file access |
INTELLIMATCH_UI_URL |
If using UI tools | IntelliMatch web app URL |
INTELLIMATCH_UI_USER |
If using UI tools | Automation user |
INTELLIMATCH_UI_PASSWORD |
If using UI tools | Automation password |
ISUITE_UI_URL |
If using UI tools | iSuite web app URL |
ISUITE_UI_USER |
If using UI tools | Automation user |
ISUITE_UI_PASSWORD |
If using UI tools | Automation password |
SPECS_DIR |
No | Override specs directory (default: ./specs) |
CONNECTIONS_FILE |
No | Override connections config path |
Security Notes
- SQL injection: DB tools use parameterized queries exclusively.
{{variable}}in theparamsarray maps to a positional placeholder ($1,?,@param1) — user input is never concatenated into SQL strings. - Path traversal: File-based log tools restrict access to the
LOG_BASE_DIRdirectory. - Secrets: Connection passwords and API tokens are read from environment variables, never stored in spec files.
- Dependency hygiene: All packages are pinned to actively maintained major versions with no known CVEs at time of writing. Run
npm auditregularly.
Troubleshooting
Tools not loading: Check the server stderr output — each spec file reports success or the Zod validation error that caused it to be skipped.
DB connection errors: DB drivers (pg, mssql, mysql2) are loaded on first use. Confirm the correct driver is installed and the connection config matches your DB type.
UI automation fails: Run npx playwright install to download browser binaries. Use headless: false temporarily to watch the browser for debugging.
Duplicate tool names: If two spec files define the same name, the server warns and the last loaded definition wins. Keep tool names unique across all spec files.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.