Overview
Terra AI Interface allows your AI agents to connect to health data sources, select different types of data and make aggregations via our MCP server. It ensures that your agent only retrieves meaningful, relevant data for reasoning, without having to overfetch or bloat their context limits. The AI interface allows you to:
Request only specific, relevant data sets
Access a user’s health memory
Access aggregated stats, trends and baselines
Access semantic meaning of health data
It allows your AI models to reason over continuous health data and flag when health trends change. Through Terra AI Interface, your AI can understand a users' health baseline, sleep, activity, nutrition, daily habits, menstruation data in detail and translate raw health metrics into semantic meaning.
Why we built Terra AI Interface
We built Terra AI Interface so AI health apps can get the right data in the most token efficient and fastest way. To date, there has been no easy and secure way for AI agents to request health data directly and build features around it.
This means that you don't have to make API calls, handle webhooks, trigger backfills, store or parse data.
What Terra AI Interface includes
Terra AI interface consists of tools and resources:
Tools: These are functions that your AI agent can call to get specific health data. They can be used to query specific data and perform aggregations. Example: The sleep tool can be used to perform an analysis of a user's sleep architecture between the 12th November 2025 and 30th November 2025.
Resources: These provide context about the available data and how the data schema is structured.
Prompts: These are text based templates that help your AI agent understand how to use the tools and how to get data from our resources.
Supported AI agents
The Terra AI interface works with any system that supports an MCP, including:
Claude desktop
VS code
Cursor
AgentGPT
OpenAI Assistants API
Anthropic Messages API
Google Gemini API
All custom AI applications that can connect to MCPs
Last updated
Was this helpful?