ForkMate

Effortless calorie tracking for people who train — just tell your AI what you ate.

ForkMate is an MCP server published by ai.forkmate that enables an AI agent to handle calorie tracking through natural language. It is designed for people who train and want to log what they eat without manual data entry. An agent using this server can accept descriptions of meals and foods spoken or typed by the user and convert that input into structured calorie records. The integration communicates over streamable HTTP transport between the host and the server. The source code is available in the public GitHub repository at https://github.com/shawnazar/forkmate. Because no specific tool schema is provided in the registry record, the exact operations an agent can invoke are not documented here, and developers should inspect the repository or rely on runtime discovery to understand the available endpoints. ForkMate is suited for fitness, nutrition, and quantified-self applications where conversational food logging is required. The server belongs to the ai.forkmate namespace and is identified in registries by the slug aiforkmate-forkmate or the canonical name ai.forkmate/forkmate. As an MCP server, it follows the Model Context Protocol, allowing compatible AI hosts to connect and delegate nutrition-logging tasks to it. Developers building training assistants or diet-tracking workflows can incorporate ForkMate to offload calorie estimation and record-keeping to a dedicated backend.

Use cases

  • Build an AI coach that logs meals for a user after a voice description.
  • Create a chatbot that asks what the user ate and records the calories without structured forms.
  • Add hands-free nutrition tracking to a training or workout assistant.
  • Prototype a quantified-self application that accepts free-text food entries.

When to use it

  • You need an MCP server that supports streamable HTTP transport.
  • Your application targets users who want to track calories while training.
  • You want to let users describe food in plain language instead of filling out forms.
  • You are comfortable reviewing the source repository to discover available capabilities.

When not to use it

  • You require documented tool schemas or an explicit tool list in the registry record.
  • You need a database-backed MCP server for querying complex relational nutrition datasets.
  • Your deployment requires a transport other than streamable HTTP.