Cursor and other assistants

LabeljoyNew version

Any AI assistant that lets you add an MCP server of your own can work in Labeljoy: give it the Labeljoy MCP server as a local server, as Connecting your AI assistant explains. Two that are often asked about follow. Before you start, turn on Allow external AI in Labeljoy: Turning on External AI.

Cursor. Cursor keeps your MCP servers in mcp.json, in the .cursor folder of your user folder (%USERPROFILE%\.cursor\mcp.json)home folder (~/.cursor/mcp.json). Create it if it isn't there, and add Labeljoy under "mcpServers", as Connecting your AI assistant shows. Then open the Agent chat and ask - the status bar shows it as Cursor while it works.

LM Studio runs an AI model on your own computer, with no account: nothing you ask leaves it. Choose a model that can use tools. Its MCP servers are in mcp.json, in the .lmstudio folder of your user folder (%USERPROFILE%\.lmstudio\mcp.json)home folder (~/.lmstudio/mcp.json): add Labeljoy under "mcpServers", as for Cursor. LM Studio lists it as mcp/labeljoy, and a chat uses it only once you turn it on in that chat's integrations - without it, the model can't reach Labeljoy, and may not say so. A model on your own computer is slower than the large ones on the internet and makes simpler labels: use it for small, precise requests - one thing at a time - and one of the assistants above for a whole label.

Any other. Look in its settings for MCP servers - or tools, connectors, local servers - and give it the four things in Connecting your AI assistant: a name, a local (stdio) server, the full path of the Labeljoy MCP server as its command, and no arguments. If it only takes a web address, it can't use Labeljoy.