Resources
MCP Server
Forkbit ships a Model Context Protocol server that exposes the API as tools any MCP-compatible LLM client (Claude Desktop, IDE assistants, agent frameworks) can call directly.
Local server, not hosted
The server runs on your machine alongside your MCP client, over stdio. There is no hosted Forkbit MCP endpoint yet, so every client needs its own FORKBIT_API_KEY.
Install
terminal
npx -y @forkbit/mcpThe process exits immediately with a clear error if FORKBIT_API_KEY is missing. Set FORKBIT_BASE_URL to point the server at a local or staging API instead of production.
Configure your client
claude_desktop_config.json
{
"mcpServers": {
"forkbit": {
"command": "npx",
"args": ["-y", "@forkbit/mcp"],
"env": { "FORKBIT_API_KEY": "fkb_live_xxxxxxxxxxxx" }
}
}
}Tools
Every tool description below is guidance the model itself reads at call time: credit cost, typical latency, and when to prefer a sibling tool instead.
| Tool | What it does | Credit cost |
|---|---|---|
search_foods(query, page?, limit?, category?) | Search by food name. Returns lightweight summaries. Use get_food or batch_get_foods for full nutrient detail. | 1 / call |
get_food(id) | Full nutrient and portion detail for one food by id. | 1 / call |
batch_get_foods(ids) | Full detail for up to 50 ids in one call. Use this instead of calling get_food in a loop. | 1 / id |
parse_nutrition(text) | Parse a free-text meal description into matched foods with nutrition. LLM-backed. | 10 / call |
analyze_recipe(servings, ingredients) | Totals and per-serving nutrition for a full recipe. Prefer parse_nutrition for independent items that do not need recipe totals. This is the most expensive tool. | 10 / ingredient line |