External Integrations
ChatKcal is designed to be an "Intelligence Sink" — receiving structured data from external AI agents.
1. The "Magic Link" Protocol
The primary ingestion mechanism is a deep link generated by an LLM.
Structure:
https://chatkcal.app/log?meal_json=<JSON_String>
(Note: We prioritize readable, structured data over URL compression for now).
JSON Payload:
{
"mealSummary": "Homemade Beef Chili",
"emoji": "\ud83c\udf72",
"calories": 650.5,
"protein": 45.2,
"carbs": 35,
"fat": 38.5,
"notes": "High fat estimate due to generous cheddar..."
}
Note on Compatibility: The system remains backward compatible with the legacy snake_case format (
meal_summary,calories_kcal, etc.).
2. Future: MCP Support
Goal: Remove the "Copy/Paste" step entirely.
We plan to implement the Model Context Protocol (MCP) to allow local LLMs (like Claude Desktop) to connect directly to ChatKcal as a tool resource.
- Tool:
add_meal(meal_data) - Flow: The user talks to Claude, Claude calls the
add_mealtool, and ChatKcal updates instantly.