Pepesto API用MCPサーバー — レシピ（URL、プレーンテキスト、写真）を、欧州26のスーパーマーケットのリアルタイム価格に基づいた買い物かごに変換する機能をエージェントに提供します。このMCPは、ワークフローの**「レシピ → マッチング済みカート」**の工程（解析、検索、材料のSKUへのマッピング、カタログ確認）をカバーします。実際の注文確定は別のステップとなります。詳細はチェックアウトの仕組みを参照してください。
claude_desktop_config.json に以下を追加します：
{
"mcpServers": {
"pepesto": {
"command": "npx",
"args": ["-y", "@pepesto/pepesto-mcp"],
"env": { "PEPESTO_API_KEY": "pep_sk_…" }
}
}
}claude mcp add pepesto -e PEPESTO_API_KEY=pep_sk_… -- npx -y @pepesto/pepesto-mcpRelated MCP server: Swissgroceries MCP
従量課金制のクレジットパックから開始してください — https://www.pepesto.com/pricing/ を参照。
チェックアウト時に使用したメールアドレスで /link を呼び出し、APIキーを発行します。キーは一度しか表示されないため、すぐに保存してください。
curl -X POST https://s.pepesto.com/api/link \
-H "Content-Type: application/json" \
-d '{"email":"you@example.com"}'環境変数にキーを設定します：
export PEPESTO_API_KEY=pep_sk_…ツール | エンドポイント | 説明 |
|
| レシピからマッチング済みカートへの一括変換。チェックアウト用の |
|
| URL/テキスト/画像のレシピを構造化された材料 + |
|
| Pepestoの100万件以上のレシピグラフを検索します。 |
|
|
|
|
| スーパーマーケットの全SKUダンプ。明示的に要求された場合のみ使用し、結果をキャッシュしてください。 |
|
| 残りクレジットを確認します。無料です。 |
このMCPは「価格付きマッチング済みカート」までを対象としています。ユーザーが注文を完了する方法についてはチェックアウトの仕組みを参照してください。/session、/checkout、/link は意図的にラップされていません。今後の計画についてはロードマップを参照してください。
最速のパスです。1回のツール呼び出しで、マッチング済みカートとチェックアウトリンクが返されます。
ユーザー: BBC Good Foodのピザ・マルゲリータのレシピを使ってTescoのカートを作成し、さらに炭酸水とオリーブオイルを追加して。
アシスタント: [
content_urls、content_text、supermarket_domain: "tesco.com"を指定してpepesto_oneshotを使用]アシスタント: Tescoでカートがマッチングされました。チェックアウトリンク:
<redirect_url>
「何か作るものを探して」というプロンプトに対し、エージェントがまずPepestoのレシピグラフを検索し、価格付きのマッチング済み商品を返す場合です。
ユーザー: 2人用のヴィーガンパスタのレシピを探して。
アシスタント: [
query: "vegan pasta dinner for two"を指定してpepesto_suggestを使用]アシスタント: 3つの選択肢があります：Creamy Cashew Penne、Lemon Garlic Spaghetti、Roasted Veg Rigatoni。どれにしますか？
ユーザー: 最初のやつで。Coop (coop.ch) で買い物して。
アシスタント: [選択したレシピの
kg_tokenとsupermarket_domain: "coop.ch"を指定してpepesto_productsを使用]アシスタント: Coopで8つの材料がマッチングされました — カシューナッツ CHF 4.20、ペンネ CHF 1.95、... — 合計 CHF 18.40。
ユーザーがすでにレシピURLやテキストを持っている場合です。MCPはマッチング済み商品リストを返します。その後の処理はユーザー次第です。
ユーザー: これを解析して材料を抽出して: https://www.bbcgoodfood.com/recipes/pizza-margherita-4-easy-steps
アシスタント: [
recipe_urlを指定してpepesto_parseを使用]アシスタント: Pizza Margherita in 4 easy steps — 8つの材料：強力粉、インスタントドライイースト、...
ユーザー: それらをSainsbury'sの商品とマッチングして。
アシスタント: [返された
kg_tokenとsupermarket_domain: "sainsburys.co.uk"を指定してpepesto_productsを使用]アシスタント: Sainsbury'sで8つの材料がマッチングされました — 合計 £6.40。
pepesto_products を通じて複数の kg_token を一度に呼び出すと、Pepestoがレシピ間で重複する材料を統合します。
ユーザー: 今週のケトディナーを5つ計画して、Delhaizeでマッチングして。
アシスタント: [
query: "keto dinner low carb high protein"、num_to_fetch: 5を指定してpepesto_suggestを使用]アシスタント: 5つのディナー：Steak & Kale、Salmon Avocado Bowl、Zucchini Lasagna、Cauliflower Risotto、Chicken Caesar。
ユーザー: 1週間分をカバーする買い物リストを1つ作って。
アシスタント: [5つすべての
kg_tokensとsupermarket_domain: "delhaize.be"を指定してpepesto_productsを使用]アシスタント: 統合マッチング：24種類のユニークな材料（レシピ間で共有されているものあり）、Delhaizeでの合計見積もり €68.40。
同じレシピを各マーケットで pepesto_products にかけ、合計を比較します。
ユーザー: この日曜ランチのレシピを解析して。
アシスタント: [
pepesto_parseを使用]アシスタント: 了解：Italian Sunday Lunch、12種類の材料。
ユーザー: EsselungaとConadの両方のカートを見せて、比較したいから。
アシスタント: [
pepesto_productsを2回使用 —supermarket_domain: "spesaonline.esselunga.it"とsupermarket_domain: "spesaonline.conad.it"]アシスタント: Esselunga: €34.20。Conad: €31.50。Conadの方が €2.70 安く、主にパスタとトマトで差が出ています。
明示的な要求があった場合のみ実行し、積極的にキャッシュしてください — 以下の価格とベストプラクティスを参照。
ユーザー: 価格比較ダッシュボードを作っているんだ。Plus NLの全カタログを引っ張ってきて。
アシスタント: 注意 —
pepesto_catalogはここで最も高価な呼び出しです。1回だけ実行します。結果は少なくとも1日はキャッシュしてください。実行しますか？ユーザー: はい。
アシスタント: [
supermarket_domain: "plus.nl"を指定してpepesto_catalogを使用]アシスタント: カタログをダンプしました：23カテゴリー、1,847 SKU。
# | 国 | スーパーマーケット | ドメイン / ID |
2 | 🇬🇧 GB | Sainsbury's | sainsburys.co.uk |
3 | 🇬🇧 GB | ASDA | asda.com |
4 | 🇬🇧 GB | Morrisons | groceries.morrisons.com |
5 | 🇬🇧 GB | Waitrose | waitrose.com |
1 | 🇬🇧 GB | Tesco | tesco.com |
6 | 🇳🇱 NL | Albert Heijn | ah.nl |
7 | 🇳🇱 NL | Jumbo | jumbo.com |
8 | 🇳🇱 NL | Plus NL | plus.nl |
9 | 🇩🇪 DE | Rewe | shop.rewe.de |
10 | 🇨🇭 CH | Coop CH | coop.ch |
11 | 🇨🇭 CH | Migros | migros.ch |
12 | 🇨🇭 CH | Farmy | farmy.ch |
13 | 🇨🇭 CH | Aldi CH | aldi-now.ch |
14 | 🇧🇪 BE | Colruyt | colruyt.be |
15 | 🇧🇪 BE | Delhaize | delhaize.be |
16 | 🇮🇪 IE | Tesco IE | tesco.ie |
17 | 🇮🇪 IE | SuperValu | shop.supervalu.ie |
18 | 🇮🇪 IE | Dunnes | dunnesstoresgrocery.com |
19 | 🇮🇹 IT | Esselunga | spesaonline.esselunga.it |
20 | 🇮🇹 IT | Conad | spesaonline.conad.it |
21 | 🇩🇰 DK | Nemlig | nemlig.com |
22 | 🇳🇴 NO | Meny | meny.no |
23 | 🇵🇱 PL | Frisco | frisco.pl |
24 | 🇵🇱 PL | Auchan PL | zakupy.auchan.pl |
25 | 🇧🇬 BG | Bulmag | bulmag.org |
26 | 🇧🇬 BG | eBag | ebag.bg |
リストにないスーパーマーケットが必要ですか？Pepestoまでご連絡ください。
このMCPは「価格付きマッチング済みカート」までを対象としています。スーパーマーケットのウェブサイトでの注文確定を自動化するものではありません。完了までの2つの方法：
Pepestoアプリ（推奨）。 pepesto_oneshot が返す redirect_url をブラウザで開くか、pepesto_products からのマッチング済み商品リストをユーザーに渡し、Pepestoアプリで再現するよう伝えてください。ログイン、バスケット確認、支払い（一部の市場）を含むホスト型チェックアウトフローが利用可能です。
スーパーマーケットの自社サイト。 ユーザーは pepesto_products からのマッチング済み商品リストを使い、tesco.comやcoop.chなどで直接SKUを追加できます。手間はかかりますが、Pepestoアカウントは不要です。
Pepestoはシンプルな従量課金制です。実際に使用した分のみ支払い、クレジットに有効期限はありません。学生や初期段階のチーム向けの割引も提供していますので、該当する場合はご連絡ください。呼び出しごとの詳細な価格とボリュームティアは https://www.pepesto.com/pricing/ を参照してください。
クレジットを最大限に活用するためのヒント：
pepesto_credits は無料です — いつでも残高を確認できます。
pepesto_oneshot、pepesto_parse、pepesto_suggest、pepesto_products は日常的な呼び出し（レシピマッチング、週間計画、バスケット比較）であり、日常的なエージェント利用に適した価格設定です。
pepesto_catalog はスーパーマーケットの全SKUをダンプするため、最も負荷の高い呼び出しです。市場分析や価格比較ダッシュボードには最適ですが、スーパーマーケットごとに少なくとも1日は結果をキャッシュしてください。必要かどうかわからない場合は、ユースケースをお知らせください。より安価な方法をご案内します。
以下を予定しています：
pepesto_session — /session をラップし、エージェントが選択したSKUからPepesto側のチェックアウトセッションを構築できるようにします。
pepesto_checkout — /checkout をラップし、スーパーマーケットの自社サイトを操作するブラウザ自動化ループ（ログイン、バスケット追加、CAPTCHA対応など）を実現します。これが完全自律型ショッピングの最後のピースです。
ホスト型チェックアウトへの引き継ぎ — Pepestoアプリのディープリンクを構造化されたツール結果として表示し、MCPクライアントがURLではなくボタンとしてレンダリングできるようにします。
これらが必要な場合はお知らせください — 優先順位を上げます。
git clone https://github.com/pepesto-solutions/pepesto-mcp.git
cd pepesto-mcp
npm install
npm run build
npm test
npm run test:coverageローカルビルドに対してインスペクターを実行します：
PEPESTO_API_KEY=pep_sk_… npm run inspectorこのリポジトリのPepesto MCPサーバーは MITライセンス の下でライセンスされています。
Dump Pepesto's full indexed catalog for a supermarket (~1-2k SKUs of common cooking ingredients, with names, prices, images, IDs). Optionally pass a webhook_url to receive incremental updates on re-index. Use only when the user has explicitly asked for a catalog dump, market analysis, or storefront build; for normal recipe-to-cart flows use pepesto_oneshot or pepesto_products instead. Cache this query aggressively, no more than one call per day per supermarket is recommended. When presenting supermarket results the user, use the product key as (external) link to the supermarket product itself. Show image if available (json property image_url, don't search for external images, skip rendering the Pepesto image if the image has webp extesion)
| Name | Required | Description | Default |
|---|---|---|---|
| supermarket_domain | Yes | Supermarket domain or ID, e.g. 'coop.ch', 'tesco.com', 'ah.nl'. See README for the full list. | |
| webhook_url | No | URL to POST incremental catalog updates to. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations provided, so description carries full burden. It discloses the optional webhook for incremental updates, caching advice, and image handling. Does not explicitly mention read-only nature or auth requirements, but covers key behaviors.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Description is dense with useful information, front-loaded with purpose and core details. Slightly long but every sentence adds value. Could be slightly more concise, but well structured.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Covers return content (names, prices, images, IDs), caching, image handling, and presentation guidance. Lacks details on error handling or webhook format, but adequate for a dump tool.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100%, baseline 3. Description adds meaning: explains webhook_url is for incremental updates, and gives example for supermarket_domain. Adds value beyond schema.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool dumps the full catalog for a supermarket with specific details (1-2k SKUs, names, prices, images, IDs). It distinguishes from siblings by mentioning alternative tools for normal recipe-to-cart flows.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Explicitly states when to use (catalog dump, market analysis, storefront build) and when not (use pepesto_oneshot or pepesto_products for normal flows). Also provides caching recommendations.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Return the remaining API credits on the configured Pepesto API key.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations provided, so the description carries full burden. It clearly states the tool returns credits with no side effects, but lacks detail on error conditions (e.g., missing key). Still sufficient for simple read operation.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Single sentence, no superfluous words. Every word contributes meaning. Perfectly concise.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
No output schema, so description should explain return value. It states returns remaining credits, but lacks format (e.g., number, string). Adequate for simple tool but could be more precise.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Input schema has zero parameters, so schema coverage is 100%. Description adds no param info, which is appropriate. Baseline score of 4 applies as per guidelines.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states that the tool returns the remaining API credits, distinguishing it from sibling tools like catalog or oneshot. The verb 'Return' and resource 'remaining API credits' are specific and unambiguous.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No explicit guidance on when to use this tool versus alternatives. However, due to its specific function, usage is implied when checking API credits. Missing when-not-to-use or prerequisites.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
One-shot: turn recipe URLs, free-form text, and/or an image into a ready-to-checkout cart for a chosen European supermarket. Returns a redirect_url that opens the Pepesto checkout UI for the user to verify and pay. Internally runs parse + products + session (not exposed to agents) with Pepesto's heuristics. Use this when you want the simplest end-to-end flow; use pepesto_parse + pepesto_products + session (not exposed to agents) for finer control.
| Name | Required | Description | Default |
|---|---|---|---|
| content_urls | No | Recipe URLs to parse and shop. | |
| content_text | No | Free-form shopping list or extra items to include. | |
| content_image | No | Base64-encoded recipe image. | |
| supermarket_domain | No | Supermarket domain or ID, e.g. 'coop.ch', 'tesco.com', 'ah.nl'. See README for the full list. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries full burden for behavioral disclosure. It mentions internal steps (parse + products + session) and that it returns a redirect_url, but does not cover error handling, idempotency, authentication needs, or potential side effects beyond cart creation. While the core behavior is described, more detail would improve transparency.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is four sentences, each serving a purpose: purpose, output, internal process, usage guidance. It is front-loaded with the main action and efficiently conveys all needed information without redundancy.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the tool's complexity (multiple input types, single output) and the lack of output schema or annotations, the description provides adequate context: what it does, what it returns, and when to use alternatives. However, it could be more complete by mentioning possible failure modes or prerequisites, but overall it is sufficient for agent decision-making.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
All four parameters are described in the input schema (100% coverage). The description loosely maps inputs to parameters (recipe URLs, free-form text, image) and gives an example for supermarket_domain, but this adds only minor context beyond the schema. Baseline 3 is appropriate.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's purpose: turning recipe URLs, free-form text, and/or an image into a ready-to-checkout cart for a European supermarket. It specifies the output (redirect_url) and distinguishes itself from siblings by noting it internally runs parse+products+session, and recommends pepesto_parse+pepesto_products for finer control.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description explicitly tells when to use this tool ('simplest end-to-end flow') and when to use alternatives ('pepesto_parse + pepesto_products + session for finer control'). This provides clear guidance for the agent to choose appropriately.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Parse a recipe from a URL, free-form text, or image into a structured object: title, ingredients, nutrition, instructions, and a KgToken you can pass to pepesto_products to build a real cart. Once the response is returned, show recipe title, image if available (json property image_url, don't search for external images, skip rendering the Pepesto image if the image has webp extesion), ingredients, steps, nutrition summary, allergens clearly marked, and portions/servings if available. Don't show kg_token, but mark and save it for the next steps (e.g., /products call).
| Name | Required | Description | Default |
|---|---|---|---|
| recipe_url | No | Publicly crawlable recipe URL. | |
| recipe_text | No | Free-form recipe text. | |
| recipe_image | No | Base64-encoded recipe image. | |
| locale | No | BCP 47 locale, e.g. 'en-GB', 'de-CH'. | |
| generate_image | No | Whether to generate a shareable image of the recipe. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations provided; description gives response handling instructions but does not disclose behavioral traits like authorization, rate limits, or error scenarios.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Description is verbose and includes agent instructions (show/hide fields) that could be separate; not concise.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Explains output fields and KgToken but lacks detail on when to use each input type and does not cover return structure.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100%; description adds context linking input types to output but does not detail each parameter's usage comprehensively.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
Clearly states verb 'Parse' and resource 'recipe', and enumerates output fields. However, no explicit differentiation from sibling tools.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Describes what to do after parsing (show fields, save kg_token) but lacks guidance on when to use this tool vs alternatives or when not to use it.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Turn a free-form shopping list into a deferred deep link to the Pepesto mobile app, returned as a redirect_url. This is a PUBLIC endpoint: it is FREE to the API client, needs no API key, and returns instantly. Parsing and product matching happen lazily, only after the user opens the link — and the USER (not the API client) is charged when they proceed to checkout in the app. If the app isn't installed, the user is sent to the app store first and the shopping list is preserved until the app opens. This is an end-user / agent-facing handoff (e.g. a person chatting in Claude Desktop who wants to finish shopping on their phone), not a developer-integration endpoint. Choose pepesto_predirect when the cost should fall on the end user and a deferred deep link is acceptable. Choose pepesto_oneshot instead when the client wants the basket matched up front (with prices) and is willing to pay for the matching. PRESENTATION (important): the tool's text output is ready-to-show Markdown — a single tappable, labeled link plus a one-line caption. Surface it to the user exactly as returned; do NOT also paste the long raw URL as plain text. You may add one short sentence telling them to open it on their phone (on a computer, opening it shows a QR code to scan).
| Name | Required | Description | Default |
|---|---|---|---|
| shopping_list | Yes | Free-form shopping list. May contain multiple newline-separated lines, e.g. '2 avocados\n1 loaf of bread\n500 g tomatoes'. | |
| locale | No | User's locale, e.g. 'de-DE'. Optional. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Despite no annotations, the description comprehensively discloses behavioral traits: it is free, requires no API key, returns instantly, parsing occurs lazily after link opening, the user is charged (not the API client), and behavior when app is not installed (app store redirect with preserved list).
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is lengthy but well-structured, front-loading the core function and then adding usage guidance, behavioral details, and presentation instructions. Every sentence adds value, though it could be slightly more concise.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Without an output schema, the description fully explains the return format (Markdown with a tappable link and caption) and instructs the agent on how to present it to the user. It also covers edge cases (app not installed) and provides a complete picture of the tool's behavior.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100% with descriptions for both parameters. The description adds minimal extra semantic beyond schema, only noting that shopping_list can be multi-line and locale is optional. This is adequate given the high coverage, but does not significantly enhance parameter understanding.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool's purpose: converting a free-form shopping list into a deferred deep link to the Pepesto mobile app. It specifies the key characteristics (free, public, no API key, lazy parsing) and distinguishes it from the sibling pepesto_oneshot tool.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description provides explicit guidance on when to use this tool versus pepesto_oneshot, based on cost allocation and timing of basket matching. It also clarifies that this is an end-user/agent-facing handoff, not a developer-integration endpoint.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Map one or more recipe KgTokens (and an optional manual shopping list) to concrete supermarket products with prices, images, or currency. Items are merged across recipes to reduce waste; multiple matches per ingredient let you (or the user) pick. Show product title, image if available (json property image_url, don't search for external images, skip rendering the Pepesto image if the image has webp extesion)product_id when available linking to an (external) supermarket page (open in a new tab), price, ProductClassification (is_bio, is_frozen, is_substitution) tags. PricePromotion shows if the item is currently on promotion and what's current promo_percentage
| Name | Required | Description | Default |
|---|---|---|---|
| recipe_kg_tokens | Yes | KgTokens from pepesto_parse or pepesto_suggest. | |
| supermarket_domain | Yes | Supermarket domain or ID, e.g. 'coop.ch', 'tesco.com', 'ah.nl'. See README for the full list. | |
| manual_shopping_list | No | Free-text extra items to add (e.g. 'milk, bananas, kitchen towel'). |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description discloses key behavioral traits: merging items across recipes, multiple matches per ingredient for user selection, output fields (title, image, price, classification, promotion), and specific rendering instructions (skip webp images). However, it does not mention idempotency, rate limits, or authorization requirements.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single paragraph that covers all necessary details without excessive length. It could benefit from bullet points for output fields, but the information is front-loaded with the main purpose in the first sentence.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the absence of an output schema, the description thoroughly explains the return values (title, image, product_id, price, classification, promotion) and the merging/user-picking behavior. It provides sufficient detail for an agent to understand the tool's function and output.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100%, so baseline is 3. The description adds only contextual behavior (merging, picking) but does not enhance parameter semantics beyond what the schema already provides (e.g., examples, source hints).
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description uses a specific verb 'Map' and clearly states the resource transformation: KgTokens to supermarket products. It distinguishes itself from sibling tools that handle parsing (pepesto_parse) or suggestion (pepesto_suggest) by focusing on product mapping.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies the tool is used after obtaining KgTokens from pepesto_parse or pepesto_suggest, but it does not provide explicit guidance on when to use this tool versus alternatives, nor does it mention any prerequisites or exclusions.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Search Pepesto's recipe graph (1M+ recipes) by free-text query and optional filters (cuisine, dietary tags, ingredients to include/avoid, time, servings). Each result includes a KgToken you can pass to pepesto_products. Returned images are licensed for display in your app or website without attribution. Show recipe title, image if available (json property image_url, don't search for external images, skip rendering the Pepesto image if the image has webp extesion), ingredients, steps, nutrition summary, allergens clearly marked, and portions/servings if available. Don't show kg_token, but mark and save it for the next steps (e.g., /products call).
| Name | Required | Description | Default |
|---|---|---|---|
| query | Yes | Free-text query that may include cuisine, dietary tags, ingredients to include or avoid, time constraints, and servings, e.g. 'vegan keto dinner low on carb for two'. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations, the description carries the full burden. It discloses that images are licensed for display, that kg_token should be saved for the next step, and provides display instructions. However, it does not mention whether the operation is read-only, auth needs, or rate limits.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is long and includes instructions for how to handle results, which could be more concise. It mixes tool functionality with behavioral instructions for the agent.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given no output schema, the description explains what the output contains (title, image, ingredients, steps, nutrition, allergens, portions, kg_token) and provides instructions on not showing certain fields. It is fairly complete for a search tool but could be more structured.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
The input schema only has one parameter 'query' with a description that already mentions free-text. The tool description adds some extra context about cuisine, dietary tags, etc., but does not add substantial meaning beyond the schema. Schema coverage is 100%, so baseline 3 is appropriate.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool searches Pepesto's recipe graph by free-text query and optional filters, and mentions it returns results with various details. However, it does not differentiate from siblings like pepesto_catalog or pepesto_oneshot.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
The description implies when to use (searching recipes with filters) and hints at a workflow using the returned KgToken for pepesto_products. But it lacks explicit guidance on when not to use or comparisons to sibling tools.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v1.7.0pepesto_predirectv1.6.0pepesto_catalogpepesto_creditspepesto_oneshotpepesto_parsepepesto_productspepesto_suggestScored across 7 tools
Each tool has a distinct role in the recipe-to-cart workflow, but pepesto_oneshot and pepesto_predirect both produce redirect URLs from free-form lists, and pepesto_parse/pepesto_suggest both return KgTokens with similar recipe presentation. The descriptions differentiate them well enough, so this is only a minor ambiguity.
All tools share the pepesto_ prefix, which is helpful, but the naming convention is mixed: some are action verbs (parse, suggest), some are resource nouns (credits, products, catalog), and some are coined terms (oneshot, predirect). The names are readable but not predictable from a single pattern.
Seven tools is well-scoped for the server's purpose. Each tool covers a distinct step or entry point in the recipe-to-cart workflow, with no obvious bloat or thinness.
The surface covers recipe discovery, parsing, product mapping, catalog access, and end-to-end checkout flows. The main gap is that no separate session/checkout tool is exposed for fine-grained flows after pepesto_products, though pepesto_oneshot covers the common end-to-end path.
Hosted MCP for e-commerce: live product catalog, stock, and pricing for AI agents.
Unlock the power of food transparency with our Open Food Facts MCP server. Easily look up any food
Hosted MCP server to manage a restaurant menu from AI agents - 39 tools over the DuckHub API.
MCP Server for an Agent Task Marketplace
