MCP tools
Expose your app's functions as tools AI assistants can call.
MCP (the Model Context Protocol) is how AI assistants like Claude call tools in the outside world. In Pikku Studio Cloud, any function in your app can be exposed as an MCP tool — which means an AI assistant can look things up in your product, create records, or kick off work, using the same checked functions the rest of your app runs on.
Labeled diagram: an "AI assistant" box on the left connected by an arrow labeled "MCP" to "Your app's functions" on the right, with a small padlock on the arrow. Caption: "Assistants call the functions you chose to expose — nothing else."
What it's for
Say you run a shop. Expose a lookUpOrder function as an MCP tool, and a support agent's AI assistant can answer "where is order #4127?" by calling your app directly — no copy-pasting, no separate integration project.
Example: an AI agent handling support chat calls your lookUpOrder and startRefund tools, and every call runs through the same permission checks as your own screens.
Only what you choose, always typed
Two things keep this safe:
- You pick the tools. Nothing is exposed unless you make it a tool. The rest of your app is invisible to assistants.
- Every tool is typed. Pikku Studio Cloud knows exactly what data each tool accepts and returns, so an assistant can't send junk or read something it shouldn't — a malformed call is rejected before your code runs.
An MCP tool is just another wire into a function you already have. Same code, same permissions, same logs — just a doorway shaped for AI assistants instead of browsers.
Where they show up
MCP tools are listed in the Wires panel alongside your HTTP routes and channels, so the single view of "every way into my app" includes what AI assistants can reach. Calls made by assistants appear in your logs like any other request.