# Watson Orchestrate [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/markusvankempen/wxo-builder-mcp-server  
**GitHub Stars:** 0  
**npm Downloads (last month):** 54  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/watson-orchestrate

## Description
Manage IBM Watson Orchestrate tools, agents, connections from Cursor, Copilot, Claude.

## Tools
Capabilities this server exposes over MCP:

- **list_skills**
- **list_tools_with_connections**
- **list_standard_tools**
- **get_skill**
- **delete_skill**
- **deploy_skill**
- **deploy_tool_from_url**
- **create_python_tool_from_tool_spec_json**
- **tool-spec.json**
- **python_code**
- **requirements**
- **create_python_tool_and_upload**
- **tool_spec**
- **python_filename**
- **create_tool_and_assign_to_agent**
- **agent_id**
- **agent_name**
- **WO_AGENT_IDs**
- **assign_tool_to_agent**
- **update_skill**
- **copy_skill**
- **new_name**
- **execute_tool**

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "watson-orchestrate": {
    "command": "npx",
    "args": ["-y","wxo-builder-mcp-server"],
    "env": {
      "WO_API_KEY": "",
      "WO_INSTANCE_URL": ""
    }
  }
}
```

**Requires environment variables:** `WO_API_KEY`, `WO_INSTANCE_URL` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What the Watson Orchestrate MCP server does

The Watson Orchestrate MCP server exposes IBM Watson Orchestrate administration and execution capabilities through MCP tools. It is intended for AI clients that need to inspect or modify Watson Orchestrate resources without directly implementing the platform's REST calls.

Available operations cover skills, tools, connections, and agents. The server can list skills, standard tools, and tools together with their connections. It can retrieve, update, copy, deploy, or delete skills, and it can execute tools through the connected Watson Orchestrate instance.

Tool creation is also supported. Inputs include a tool specification, Python source code, requirements, and a Python filename. The listed creation operations support building Python tools from a specification or uploading a Python tool, including creating a tool and assigning it to an agent.

## How it works

An MCP client communicates with the local server using MCP tool calls. The server then acts as an HTTP client and forwards requests to the configured Watson Orchestrate instance. Watson Orchestrate does not connect back to the MCP process.

The package is distributed through npm as `wxo-builder-mcp-server`. A client configuration starts it with `npx -y wxo-builder-mcp-server`. Cursor uses an `mcpServers` configuration section, while VS Code Copilot uses `servers` with a `stdio` entry. The repository also provides example configurations for Claude Desktop, Antigravity, and Windsurf.

## Setup and configuration

Install the package with npm or invoke it directly through `npx`. Set these required environment variables in the MCP client configuration or an environment file:

- `WO_API_KEY`: IBM Cloud API key used to access Watson Orchestrate.
- `WO_INSTANCE_URL`: URL of the Watson Orchestrate instance, such as an `orchestrate.ibm.com` host.

Agent-oriented tools can use a default agent when the request does not specify an agent. Set `WO_AGENT_IDs` to a comma-separated list, or set the backward-compatible `WO_AGENT_ID` variable for a single default. When multiple IDs are supplied, the first ID is used by default.

The Watson Orchestrate MCP server is configured per client. The README includes copy-ready examples for Cursor, VS Code, Claude Desktop, Antigravity, and Windsurf. Cursor also supports a one-click installation link, after which the API key and instance URL must be supplied in the MCP settings.

## Tools and capabilities

The exposed tool set includes:

- Discovery: `list_skills`, `list_tools_with_connections`, and `list_standard_tools`.
- Skill administration: `get_skill`, `delete_skill`, `deploy_skill`, `update_skill`, and `copy_skill`.
- Remote tool creation: `deploy_tool_from_url`, `create_python_tool_from_tool_spec_json`, and `create_python_tool_and_upload`.
- Agent assignment: `create_tool_and_assign_to_agent` and `assign_tool_to_agent`.
- Execution: `execute_tool`.

Creation and assignment inputs include tool specifications, Python code, requirements, filenames, agent IDs, agent names, and new names where applicable. The package also includes an OpenAPI specification describing the Watson Orchestrate API.

The Watson Orchestrate MCP server can be paired with the WxO Builder VS Code extension. The extension is described as providing visual editing, drag-and-drop agent editing, and local or remote testing, while the MCP package provides AI-driven access to the Watson Orchestrate capabilities.

## Limitations and notes

This package requires access to a Watson Orchestrate instance and a valid API key. Its role is to proxy MCP requests to the configured REST API; it does not provide a standalone Watson Orchestrate environment.

Agent defaults are optional, and requests can provide an agent identifier or name instead. If no explicit agent is provided, the first configured default ID is used when one is available.

_Full upstream README: https://allmcps.com/mcp/watson-orchestrate/readme_