Food nutrition lookup via natural language, barcode, and image. Database of 1,000,000+ foods.
Maintain this server? Claim it
Call this server's tools from your browser. Nothing to install.
Bearer <app_id>:<app_key> — your Edamam API credentials
Requests go through mcptry because browsers can't call most MCP servers directly. Header values are sent to the server for that request only and are never stored or logged.
Tool list from our last check. Connect live to refresh it.
Identify food in a photograph and return a full nutritional breakdown. Uses Edamam's AI Vision API (beta) to detect each dish on the plate, estimate its ingredients, and calculate nutrition. Every dish is analyzed independently, so multi-dish plates return one entry per dish. Works best with clear, well-lit photos. The image is supplied as base64-encoded bytes (the API accepts only a raw image upload, not a URL), most conveniently as a data URI: "data:image/jpeg;base64,...". A caller with a remote URL must download the image itself and base64-encode the bytes. To look up nutrition from a text description instead, use get_food_nutrition.
A base64 data URI (e.g. "data:image/jpeg;base64,...") — or a bare base64 string — of a food photograph.
This tool isn't marked read-only, so it may change something on the server or in an account you connected.
Pick your client. Values in capitals, like YOUR_TOKEN, are yours to fill in.
Run in a terminal
claude mcp add --transport http food https://mcp.edamam.com/mcp/food \ --header 'Authorization: Bearer YOUR_TOKEN'
Headers
Authorization (required): Bearer <app_id>:<app_key> — your Edamam API credentialsIdentify food in a photograph and return a full nutritional breakdown. Uses Edamam's AI Vision API (beta) to detect each dish on the plate, estimate its ingredients, and calculate nutrition. Every dish is analyzed independently, so multi-dish plates return one entry per dish. Works best with clear, well-lit photos. The image is supplied as base64-encoded bytes (the API accepts only a raw image upload, not a URL), most conveniently as a data URI: "data:image/jpeg;base64,...". A caller with a remote URL must download the image itself and base64-encode the bytes. To look up nutrition from a text description instead, use get_food_nutrition.
| Argument | Type | Description |
|---|---|---|
| image* | string | A base64 data URI (e.g. "data:image/jpeg;base64,...") — or a bare base64 string — of a food photograph. |
{
"type": "object",
"required": [
"image"
],
"properties": {
"image": {
"type": "string",
"description": "A base64 data URI (e.g. \"data:image/jpeg;base64,...\") — or a bare\n base64 string — of a food photograph."
}
},
"additionalProperties": false
}Look up nutrition facts for one or more foods described in plain text. Use this tool for any question about calories, macros, or nutrients for a named food or meal. Each query is a natural-language phrase describing one food item; the API identifies the food, resolves the quantity and unit, and returns a full nutrient breakdown. If the user provides a food photograph rather than a text description, use analyze_food_image instead. Returns a list of dicts, one per query. Each dict has: `food` (name, food_id, amount, unit, weight_g, ingredients), `macronutrients` (calories, protein, carbs, fat, etc.), `micronutrients` (vitamins and minerals), `diet_flags`, `health_flags`, and `status`. Returns {"error": ..., "query": ...} for any item that cannot be identified.
| Argument | Type | Description |
|---|---|---|
| queries* | array | One natural-language phrase per food item. Supported formats: plain name ("cheddar cheese", defaults to 100 g), quantity + unit ("1 cup rice", "200g chicken breast"), or UPC/EAN barcode ("049000042566", 8-14 digits). Keep compound food names as a single entry — "carrot cake", "chicken noodle soup", and "grilled cheese sandwich" are each one food, not multiple. Only split into separate entries when the user is clearly listing multiple distinct foods — e.g. "1 croissant and 1 egg" → ["1 croissant", "1 egg"]. |
{
"type": "object",
"required": [
"queries"
],
"properties": {
"queries": {
"type": "array",
"items": {
"type": "string"
},
"description": "One natural-language phrase per food item. Supported formats:\n plain name (\"cheddar cheese\", defaults to 100 g), quantity + unit\n (\"1 cup rice\", \"200g chicken breast\"), or UPC/EAN barcode\n (\"049000042566\", 8-14 digits). Keep compound food names as a\n single entry — \"carrot cake\", \"chicken noodle soup\", and\n \"grilled cheese sandwich\" are each one food, not multiple.\n Only split into separate entries when the user is clearly\n listing multiple distinct foods —\n e.g. \"1 croissant and 1 egg\" → [\"1 croissant\", \"1 egg\"]."
}
},
"additionalProperties": false
}We read every tool's name, description and inputs the way the model will, looking for instructions hidden from you. A clean result means no warning signs, not a guarantee. Scanned Nov 18, 2026.
Insists the model reach for this tool first for a whole class of requests, ahead of whatever else could answer: common in ad and sponsored-results tools.
Fix: Say when the tool is useful ("Use for...") instead of demanding it be called first or for any request.
No readOnlyHint / destructiveHint annotations, so clients can't tell a read from a write without asking.
Fix: Add annotations to each tool, e.g. { readOnlyHint: true } for tools that only read.