get_agent
Get raw agent config (name, tools, instructions) from Watson Orchestrate. Use only when raw JSON is needed. For "what can you do" or "list capabilities", prefer invoke_agent instead — it lets the agent answer directly.
LABSCO SUMMARY
One Orchestrate agent, reachable from the editor you already work in.
There is no agent management here and no tool routing - the agent is chosen once in configuration, and everything after that is a message and a reply. That is the point: an agent that already has its instructions and tools becomes callable from Cursor, VS Code or Langflow without carrying the rest of the platform.
What it is
A small server that fronts a single Watson Orchestrate agent. The agent is fixed at configuration time; the server sends it messages and returns what it says.
What you get
- Send a message to the configured Watson Orchestrate agent and get its reply - the agent answers using its own tools and LLM
- The agent's raw configuration - name, tools and instructions - when the JSON itself is what you need
Requirements
A Watson Orchestrate instance and an agent inside it. Three values identify them - WO_API_KEY, WO_INSTANCE_URL and WO_AGENT_ID - set as environment variables or in your MCP config. It installs from npm and runs over stdio.
Setup effort
One command plus a key — npx -y wxo-agent-mcp, then supply credentials
USE CASES
01
Talking to an Orchestrate agent from your editor
Send it the question in the tool you are already in, and let it answer using the tools it was built with.
02
Checking what the agent has been told before you trust its answer
Read the agent's name, instructions and tool list as raw configuration, rather than inferring them from how it replies.
03
Working out why a reply came back wrong
Put the message that produced it next to the agent's own configuration, and see whether the instructions or the tools are the cause.
2 tools
Counted once with only startup setup. Other versions, modes or optional credentials may expose a different list.
invoke_agent
Send a message to the Watson Orchestrate agent. The agent responds using its tools and LLM. Prefer this for "what can you do", "list capabilities", "what do you do" — the agent answers in its own voice. Use get_agent only when raw config/JSON is needed.
BEFORE YOU INSTALL
What to know first
Tools you get
2 tools when we connected. The 30-plus watsonx Orchestrate tools such as list_skills and deploy_tool belong to the separate wxo-builder-mcp-server package.
Requires
- npx on PATH
- WO_API_KEY in your environment
CONTEXT COST
Light Standing cost
Rated by the size of the 2 tool definitions that ride along on every turn, not by how many there are
INSTALLATION
Connect your client
Server setup
TERMINAL
npx -y wxo-agent-mcpMCP Clients
Cursor, Antigravity
JSON
{
"mcpServers": {
"wxo-agent": {
"command": "npx",
"args": [
"-y",
"wxo-agent-mcp"
],
"env": {
"WO_API_KEY": "<YOUR_WO_API_KEY>",
"WO_INSTANCE_URL": "https://<YOUR_SUBDOMAIN>.orchestrate.ibm.com",
"WO_AGENT_ID": "<YOUR_WO_AGENT_ID>"
}
}
}
}VS Code
VS CODE · JSON
{
"servers": {
"wxo-agent": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"wxo-agent-mcp"
],
"env": {
"WO_API_KEY": "<YOUR_WO_API_KEY>",
"WO_INSTANCE_URL": "https://<YOUR_SUBDOMAIN>.orchestrate.ibm.com",
"WO_AGENT_ID": "<YOUR_WO_AGENT_ID>"
}
}
}
}
