WxO Agent MCP

io.github.markusvankempenPublished by GitHub account markusvankempenv1.0.8Updated Oct 24, 2026

Invoke a Watson Orchestrate agent via MCP. Configure with WO_AGENT_ID.

VerifiedSTDIODesktop onlyProductivity & WorkflowAI & ML

Overview

AI-generated overview

Lets an assistant send messages to one preconfigured IBM Watson Orchestrate agent and read that agent's details.

What you can do
A lightweight local MCP server that connects to a single Watson Orchestrate agent over HTTP/REST. It exposes two tools: invoke_agent, which sends a message to the configured agent and returns its reply, and get_agent, which returns the agent's name, description, tools and instructions. The agent is fixed once through environment variables, so there is no agent listing, tool management or flow handling.
When to use it
Use it when you want an assistant to chat with one specific Watson Orchestrate agent and nothing more. It suits simple question-and-answer or agent-detail lookups rather than building or managing Watson Orchestrate resources; the README points to a separate builder server for that.
Requirements
Runs locally over stdio, typically via npx from the npm package wxo-agent-mcp or from a built local copy, so Node.js is needed. Requires an IBM Cloud API key in WO_API_KEY, the Watson Orchestrate instance URL in WO_INSTANCE_URL, and an agent ID in WO_AGENT_ID (or WO_AGENT_IDs, first used). Network access to the Watson Orchestrate instance is required.
Before you install
The server needs an IBM Cloud API key supplied as WO_API_KEY, which is a credential and should be kept out of shared config files. Messages sent through invoke_agent are forwarded to the remote Watson Orchestrate agent and processed there, so anything you ask leaves the local machine. The agent may call its own tools and LLM, so replies and any resulting actions depend on how that agent is configured.

Installation

In SourceWeft

  1. Open WxO Agent MCP in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

WxO Agent MCP

[npm version] [npm downloads] [License]

[wxo-agent-mcp]

[Add MCP server to Cursor]

Simple MCP (Model Context Protocol) server that invokes a single Watson Orchestrate agent remotely. The agent is defined once via environment variables or MCP config.

Use this when you want a lightweight MCP that only chats with one agent—no tool management, no agent listing, no flows. Just invoke_agent(message) and get_agent().

Full documentation: DOCUMENTATION.md (includes screenshots for VS Code and Langflow).
Publishing: PUBLISHING.md – npm and MCP Registry procedure.

Architecture

                    ┌───────────────────────────────────────┐                    │  Cursor • VS Code • Langflow • etc.   │                    │  (MCP clients)                        │                    └──────────────────┬───────────────────┘                                       │ stdio / JSON-RPC                                       ▼                    ┌───────────────────────────────────────┐                    │  wxo-agent-mcp                        │                    │  invoke_agent • get_agent              │                    └──────────────────┬───────────────────┘                                       │ HTTP / REST                                       ▼                    ┌───────────────────────────────────────┐                    │  Watson Orchestrate                   │                    │  agent + tools + LLM                   │                    └───────────────────────────────────────┘

Tools

ToolDescription
invoke_agentSend a message to the configured Watson Orchestrate agent. The agent responds using its tools and LLM.
get_agentGet details of the configured agent (name, description, tools, instructions).

Configuration

Set these in .env or your MCP client config (e.g. Cursor mcp.json):

env
WO_API_KEY=your-ibm-cloud-api-keyWO_INSTANCE_URL=https://your-instance-id.orchestrate.ibm.comWO_AGENT_ID=your-agent-id# Or WO_AGENT_IDs=id1,id2  (first is used)

Quick Start

bash
npm installnpm run buildcp .env.example .env# Edit .env with your credentials and agent IDWO_API_KEY=... WO_INSTANCE_URL=... WO_AGENT_ID=... node dist/index.js

Verify

bash
# Default question ("Hello, who are you?")WO_API_KEY=... WO_INSTANCE_URL=... WO_AGENT_IDs=... npm run test:verify
# Custom questionnpm run test:verify -- -ask "What is the weather in Amsterdam?"

Runs get_agent and invoke_agent to confirm connectivity.

Test in VS Code

  1. Open the wxo-agent-mcp folder in VS Code.
  2. Run npm run build.
  3. .vscode/mcp.json registers the MCP server as wxo-agent and loads .env.
  4. Open Copilot Chat (Ctrl+Shift+I) and ask: Use invoke_agent to ask: What is the weather in Amsterdam?
    For "what can you do", use Use invoke_agent to ask: What can you do? to avoid Copilot calling get_agent first.
  5. Or run npm run test:verify from the terminal (Ctrl+`).

See DOCUMENTATION.md for more prompts and setup.

Question examples: Full list →

invoke_agentget_agent
What is the weather in Amsterdam?Use get_agent to show agent details
Tell me a dad jokeWhat tools does my agent have?
What time is it in Tokyo?
What can you help me with?

Langflow: Add an MCP Tools component, choose STDIO, set Command=node, Args=["/path/to/dist/index.js"], and env vars. See DOCUMENTATION.md.

MCP Client Configuration

One-click install (Cursor Directory)

Add to Cursor — Opens the Cursor Directory page. After adding, set WO_API_KEY, WO_INSTANCE_URL, and WO_AGENT_ID in Cursor MCP settings.

For cursor.directory/mcp/new submission:

FieldValue
Cursor Deep Linkcursor://anysphere.cursor-deeplink/mcp/install?name=WxO%20Agent%20MCP&config=eyJjb21tYW5kIjoibnB4IiwiYXJncyI6WyIteSIsInd4by1hZ2VudC1tY3AiXSwiZW52Ijp7IldPX0FQSV9LRVkiOiJ5b3VyLWFwaS1rZXkiLCJXT19JTlNUQU5DRV9VUkwiOiJodHRwczovL3lvdXItaW5zdGFuY2Uub3JjaGVzdHJhdGUuaWJtLmNvbSIsIldPX0FHRU5UX0lEIjoieW91ci1hZ2VudC1pZCJ9fQ==
Install instructionshttps://github.com/markusvankempen/wxo-agent-mcp#mcp-client-configuration

Cursor (.cursor/mcp.json)

json
{    "mcpServers": {        "wxo-agent": {            "command": "npx",            "args": ["-y", "wxo-agent-mcp"],            "env": {                "WO_API_KEY": "your-api-key",                "WO_INSTANCE_URL": "https://xxx.orchestrate.ibm.com",                "WO_AGENT_ID": "your-agent-id"            }        }    }}

VS Code Copilot (.vscode/mcp.json)

The server is named wxo-agent. Use envFile to load .env:

json
{    "servers": {        "wxo-agent": {            "type": "stdio",            "command": "node",            "args": ["${workspaceFolder}/dist/index.js"],            "envFile": "${workspaceFolder}/.env"        }    }}

For npx (after publishing), use "command": "npx", "args": ["-y", "wxo-agent-mcp"], and add env with your credentials.

Antigravity

Same config as Cursor. Copy examples/antigravity-mcp.json or add to your MCP config:

json
{    "mcpServers": {        "wxo-agent": {            "command": "npx",            "args": ["-y", "wxo-agent-mcp"],            "env": {                "WO_API_KEY": "your-api-key",                "WO_INSTANCE_URL": "https://xxx.orchestrate.ibm.com",                "WO_AGENT_ID": "your-agent-id"            }        }    }}

vs wxo-builder-mcp-server

wxo-agent-mcpwxo-builder-mcp-server
PurposeInvoke one agentFull dev toolkit (tools, agents, connections, flows)
AgentSingle WO_AGENT_IDMultiple agents, WO_AGENT_IDs
Toolsinvoke_agent, get_agent30+ tools (list_skills, deploy_tool, etc.)
Use caseChat with a specific agentBuild and manage Watson Orchestrate resources

Links

License

Apache-2.0

Source: README.md at commit 296ff0b

Tools

0
SourceWeft has not started this package in its sandbox yet, so its tools are not known.

Version history

1
  1. v1.0.8LatestSep 16, 2026

Score

40% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts

    Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.

  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts

    Declare a license in the repository or the registry entry.