Using with AI Agents
Any AI agent that can write code (such as Claude Code, Codex, Cursor, etc.) can drive Margrete RPC directly. This page explains how to give your agent enough context to get more accurate code suggestions.
Providing API Context
Option 1: Use the Source Code
Giving the agent the margrete_rpc source code yields the most accurate API information — typically more reliable and complete than the documentation.
Ask your coding agent to run the following command to find the package path and read the installed source:
python -c "import margrete_rpc, os; print(os.path.dirname(margrete_rpc.__file__))"git clone https://github.com/Foahh/margrete-rpcAdd the src/margrete_rpc/ directory to the agent's context.
Option 2: Use the Documentation
Every page on this site provides a Markdown version, which is well-suited for asking AI questions on the web:
- Full-site index —
/llms.txtlists all pages and their links, making it a good starting point. - Individual pages — Click the Copy Markdown or Open button at the top of a page, or append
.mdto the page URL. - API reference — All classes and methods are listed under margrete_rpc.
Provide https://mg.foahh.com/llms.txt to the agent; it can start from there and fetch whichever
documentation pages it needs.
Example Workflow
Using Claude Code as an example:
-
Launch Claude Code in your project directory.
-
Ask Claude to locate the installed package path and read the source:
Find the installation path of the margrete_rpc package and read its source code. -
Describe what you want to accomplish, for example:
Write a script that finds all Tap objects in the chart with a time value less than beat 100 and deletes them.
The agent will automatically figure out what needs to be done based on the source code.