webMethods Integration Server MCP Server

An MCP (Model Context Protocol) server that gives AI assistants full control over webMethods Integration Server -- replacing webMethods Designer for most development and administration tasks.

Single ~4MB binary, no runtime dependencies, works with any remote IS instance via pure HTTP APIs.

Compatible with any MCP client: IBM Bob, Claude Code, Claude Desktop, Cursor, Windsurf, etc.

What can it do?

Area	What the AI can do	Tools
Flow services	Create services with full logic (INVOKE, MAP, BRANCH, LOOP), signatures, test them	5
Adapter services	Browse database tables/columns interactively, create Select/Insert/CustomSQL services -- like Designer	5
JDBC/SAP/OPC adapters	Create connections, listeners, notifications, query metadata	20
Kafka streaming	Create Kafka connections, event specifications, triggers	15
JMS messaging	Create JMS connections and triggers via JNDI providers	14
MQTT messaging	Create MQTT connections and triggers	12
Scheduler	Create/manage scheduled tasks (one-time, repeating, complex)	10
Users & security	Create users/groups, manage ACLs, keystores, OAuth clients	24
Server admin	Health, stats, thread dumps, logs, extended settings, license info	11
Packages & namespace	Full CRUD on packages, folders, nodes, document types	10
Ports	HTTP/FTP/FilePolling/WebSocket listener management	8
JDBC pools	Connection pool CRUD, driver management	8
Global variables	Configuration variable management	5
Remote servers	IS-to-IS remote server aliases	4
Web services	REST/SOAP endpoints, OpenAPI generation	8
Auditing	Audit logger management	5

167 tools + 9 interactive prompts in total.

Interactive setup wizards (prompts)

The AI can guide you step-by-step through setting up:

Prompt	What it does
setup_jdbc_connection	Walks through JDBC connection + adapter service creation, browses tables/columns interactively
setup_kafka_streaming	Sets up Kafka connection alias + event specification
setup_jms_connection	Creates JNDI provider + JMS connection (guides JAR installation)
setup_mqtt_connection	Sets up MQTT connection + trigger subscription
setup_sap_connection	Configures SAP adapter connection
setup_scheduled_task	Schedules a service for execution
setup_rest_api	Exposes services via OpenAPI or imports an API spec
setup_user_management	Creates users, groups, ACLs
setup_oauth	Registers OAuth clients and scopes

Quick Start

1. Install

# From crates.io
cargo install wm-mcp-server

# Or download a pre-built binary from Releases
# https://github.com/cpoder/wm-mcp-server/releases

Binaries available for: Linux (x86_64, aarch64), macOS (x86_64, Apple Silicon), Windows (x86_64).

2. Configure your MCP client

Add to .mcp.json in your project directory:

{
  "mcpServers": {
    "webmethods-is": {
      "command": "wm-mcp-server",
      "env": {
        "WM_IS_URL": "http://your-is-host:5555",
        "WM_IS_USER": "Administrator",
        "WM_IS_PASSWORD": "your-password"
      }
    }
  }
}

Multiple IS instances

Create a config file (e.g., wm-instances.json):

{
  "instances": {
    "dev": { "url": "http://dev-host:5555", "user": "Administrator", "password": "manage" },
    "prod": { "url": "http://prod-host:5555", "user": "Administrator", "password": "secret", "timeout": 60 }
  },
  "default": "dev"
}

{
  "mcpServers": {
    "webmethods-is": {
      "command": "wm-mcp-server",
      "env": { "WM_CONFIG": "/path/to/wm-instances.json" }
    }
  }
}

Every tool accepts an optional instance parameter. Use list_instances to see what's configured.

3. Use

Natural language examples:

"Create a flow service that greets the user by name"

The AI creates the package, folders, service with full flow logic, and tests it.

"Set up a JDBC connection to my SQL Server database and create a Select service for the orders table"

The AI uses the setup_jdbc_connection prompt to walk you through credentials, then browses your database tables/columns interactively (like Designer) to build the adapter service.

"Schedule pub.flow:debugLog to run every 5 minutes"

The AI uses scheduler_task_add with a repeating interval.

"Create an MQTT connection to my Mosquitto broker and subscribe to sensor/temperature"

The AI creates the connection, enables it, verifies connectivity, and sets up a trigger.

How it works

AI Assistant ──MCP (stdio)──> wm-mcp-server (Rust) ──HTTP/JSON──> webMethods IS (port 5555)

All operations use IS built-in HTTP services. No filesystem access, no SSH, no agents on the IS host.

Flow services are created via wm.server.ns/putNode which accepts the full flow tree as JSON -- the same FlowElement / Values serialization used internally by the IS runtime.

Requirements

• webMethods Integration Server 11.x with HTTP access enabled
• IS admin credentials (Administrator or equivalent)
• For JMS: provider client JARs in WmART/code/jars/static/ + IS restart
• For Kafka streaming: Kafka client JARs in WmStreaming/code/jars/static/

License

MIT