🎮 Sofort ausprobieren — keine Registrierung erforderlich. Klicke oben auf "Try in Browser" → lass BEIDE Felder für Umgebungsvariablen LEER → klicke auf Start Inspector. Der Demo-Modus greift auf echte Wikipedia- / httpbin- / Live-FX-APIs zu. Kein Konto, keine Karte, kein API-Key.
📦 v0.5.0 Umbenennung: Das npm-Paket heißt jetzt
pay-per-call-mcp(vormalslemon-cake-mcp). Der alte Name funktioniert weiterhin als dünner Wrapper, aber neue Konfigurationen solltennpx -y pay-per-call-mcpverwenden.
Gib deinem KI-Agenten eine Wallet. Pay-per-call USDC-Zahlungen für jede HTTP-API — direkt aus Claude Desktop, Cursor, Cline oder jedem anderen MCP-Client. Kein Mensch im Prozess, keine API-spezifischen Registrierungen, kein Jonglieren mit API-Keys.
Mit dem MCP-Server von LemonCake können Sie über MCP-kompatible Clients wie Claude Desktop, Cursor oder Cline kostenpflichtige APIs ohne menschliches Eingreifen mit USDC aufrufen.
Englisch ↓ Quickstart · Tools · Use Cases · Kompatibilität
Für die Nutzung des MCP-Servers sind ein LemonCake-Konto und ein USDC-Guthaben erforderlich.
Kostenloses Konto erstellen — Erledigt mit nur einer E-Mail-Adresse.
Guthaben aufladen — Mindestens $5 USDC oder JPYC (Billing).
Buyer JWT kopieren — Über Dashboard → API Keys.
In der unten stehenden claude_desktop_config.json konfigurieren.
📚 Details: Quickstart-Dokumentation
Related MCP server: pay-mcp
Füge dies zu ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) oder %APPDATA%\Claude\claude_desktop_config.json (Windows) hinzu:
{
"mcpServers": {
"pay-per-call": {
"command": "npx",
"args": ["-y", "pay-per-call-mcp"],
"env": {
"LEMON_CAKE_BUYER_JWT": "eyJhbGci..."
}
}
}
}Starte Claude Desktop neu, und die LemonCake-Tools erscheinen unter dem 🔨 Werkzeug-Symbol.
Verwende analog dazu den Server-Startbefehl npx -y pay-per-call-mcp und setze LEMON_CAKE_BUYER_JWT als Umgebungsvariable.
Node.js-Anforderung: v20 oder höher
Tool-Name | Zweck | Hauptparameter |
| Ersteinrichtungs-Guide (gibt Anweisungen zur Kontoerstellung/Aufladung) | — |
| Liste der im LemonCake-Marktplatz verfügbaren kostenpflichtigen APIs |
|
| Aufruf eines Dienstes mit Bezahlung via Pay Token |
|
| Aktuelles USDC-Guthaben und KYA-Limit abrufen | — |
| Überprüfung der Registrierungsnummer für qualifizierte Rechnungen über die NTA-API |
|
| Nutzungsstatistiken und Abrechnungshistorie pro Dienst | — |
Alle Argument-Schemata sind über den MCP Inspector oder tools/list abrufbar.
Wenn du den Server startest, ohne LEMON_CAKE_BUYER_JWT / LEMON_CAKE_PAY_TOKEN zu setzen, wechselt er automatisch in den DEMO-MODUS. Folgendes funktioniert ohne Registrierung:
list_services → Zeigt den echten Marktplatz + 3 Demos: demo_search (Wikipedia), demo_echo (httpbin), demo_fx (open.er-api).
call_service → demo_*-Dienste rufen echte kostenlose APIs auf und liefern Rohdaten (keine Kosten, keine Authentifizierung).
check_balance → Gibt ein Mock-Guthaben von $1.00 zurück (mode: "demo").
check_tax / get_service_stats → Wie gewohnt (erfordern ohnehin keine Authentifizierung).
→ Auch im Glama Inspector oder in der npm-Testumgebung kannst du das Verhalten ohne Konfiguration prüfen. Wenn du echte kostenpflichtige APIs aufrufen möchtest, setze einfach LEMON_CAKE_PAY_TOKEN.
Dieser Server unterstützt die Prompts-Fähigkeit von MCP. Die folgenden Presets erscheinen im "Prompt Picker" von Glama Inspector / Claude Desktop / Cursor und können mit einem Klick ausgeführt werden:
Prompt-Name | Inhalt | Auth |
🎮 | Demo-Modus: setup → list_services → demo_search → demo_fx in einem Durchgang | Nein |
🛍 | Listet alle Dienste des echten Marktplatzes und empfiehlt Top 3 nach Zweck | Nein |
🇯🇵 | Validierung der Steuernummer für qualifizierte Rechnungen via NTA + Quellensteuerprüfung | Nein |
💰 | Demonstriert Budgetverwaltung: check_balance → call_service → check_balance | demo OK / für echte Dienste PAY_TOKEN |
🔄 | Vergleicht dieselbe Abfrage zwischen demo_search und echtem Serper | demo OK / für Vergleich PAY_TOKEN |
🏯 | Japanische Unternehmensrecherche kombiniert mit gBizINFO + NTA + e-Gov | PAY_TOKEN (Demo-Teile nicht nötig) |
→ Sobald du pay-per-call MCP (ehemals lemon-cake) in Glama / Claude Desktop aktivierst, werden diese automatisch als Vorschläge angezeigt.
In Claude Desktop:
„Rufe mit LemonCake
demo_agent_search_apifür 0,50 USDC auf und suche nach 'AI agent payments'.“
Claude führt automatisch aus:
Überprüfung des Setup-Status via setup (nur beim ersten Mal).
call_service(serviceId="demo_agent_search_api", limitUsdc="0.50", body={query:"AI agent payments"}).
Zusammenfassung und Antwort mit den Ergebnissen.
Autonome Forschungs-Agenten — Lass deinen Agenten für Premium-Suche, Scraping oder Daten-APIs bezahlen, ohne deine Kreditkartendaten preiszugeben.
Multi-API-Workflows — Ein JWT, ein Guthaben, Dutzende von Upstream-APIs. Keine Registrierung pro Anbieter oder rotierende Keys nötig.
Compliance-bewusste Ausgaben — KYA-Limits (Know-Your-Agent) begrenzen, wie viel ein Agent pro Sitzung/Tag ausgeben darf.
Japanische Steuerautomatisierung — check_tax validiert Nummern für qualifizierte Rechnungen gegen die NTA-API für steuerliche Konformität.
Idempotente Wiederholungen — idempotencyKey macht call_service sicher für Wiederholungen, ohne doppelte Abbuchungen zu riskieren.
Client | Status | Hinweise |
Claude Desktop (macOS / Windows) | ✅ | Primäres Ziel |
Cursor | ✅ | stdio-Transport |
Cline (VS Code) | ✅ | stdio-Transport |
Claude Code CLI | ✅ | stdio-Transport |
Continue.dev | ✅ | MCP-Unterstützung seit v0.9 |
Benutzerdefinierte MCP-Clients | ✅ | Jeder Client, der MCP 1.10+ über stdio spricht |
Variablenname | Erforderlich | Beschreibung |
| ✅ | Buyer JWT (erhältlich im Dashboard unter Settings → API Keys) |
| — | Pay Token JWT (erforderlich für |
| — | API-Endpunkt (Standard: |
git clone https://github.com/evidai/lemon-cake.git
cd lemon-cake/mcp-server
npm install
npm run build
npm startdocker build -t pay-per-call-mcp .
docker run --rm -i -e LEMON_CAKE_BUYER_JWT=eyJhbGci... pay-per-call-mcpDas Image wird auch für die In-Browser-Vorschau des Glama Inspectors verwendet.
MIT © LemonCake
Invoke an upstream API service through LemonCake's pay-per-call proxy. Each successful call automatically debits USDC from your wallet via the permit you signed. LemonCake never holds your USDC — the transfer is direct wallet → provider on Base.
PRECONDITIONS:
• LEMON_CAKE_PERMIT env var must be set for real services. Get one in ~30 seconds at
https://lemoncake.xyz/start/v2 (sign in with Google, sign 1 EIP-712 permit, copy the blob).
If missing, the tool returns a structured PERMIT_MISSING error with how-to-fix steps.
• DEMO MODE: any serviceId starting with demo_ works WITHOUT a permit and hits real
free upstreams (Wikipedia / httpbin / Open-Meteo / Nominatim / MyMemory / dictionaryapi /
worldtimeapi / open.er-api). 8 demos cover search / echo / fx / translate / weather /
geocode / time / dictionary — useful for Glama Inspector or new-user trial. They are
marked with mode: "demo" and incur no charge.
• serviceId must come from list_services.
BEHAVIOR: • Returns the upstream response body verbatim (JSON or text), plus the X-Charge-Id and X-Amount-Usdc headers reported by the proxy. • HTTP 402 Payment Required is returned as a normal result (NOT thrown) so the agent can autonomously stop spending when the daily $25 permit cap is exhausted. • Pass the same idempotencyKey to retry safely without double-charging. • This tool spends real money and contacts an external service — it is non-idempotent by default and has external side effects.
x402-COMPATIBLE INTERFACE:
• Successful calls include an x402Receipt field with { scheme, chain, asset, amount,
recipient, paymentIntentId, settledAt }. Same shape as on-chain x402 receipts so the
agent's payment-handling logic is portable.
• If upstream returns an x402 challenge (WWW-Authenticate: x402, X-402-* headers, or
body.x402), it's parsed into x402Challenge for the agent to reason about.
• If upstream returns 202 + Retry-After + X-Payment-Status: pending, the result is
{ status: "PAYMENT_PENDING", paymentIntentId, retryAfterMs, retryContract }.
Re-call with the same idempotencyKey to resume — no double-charge.
Returns: { status, chargeId, amountUsdc, response, x402Receipt?, x402Challenge?, hint? }
| Name | Required | Description | Default |
|---|---|---|---|
| body | No | JSON request body (only used for POST/PUT/PATCH). | |
| path | No | Sub-path on the service (e.g. "/search", "/v1/completions"). Defaults to "/". | / |
| method | No | HTTP method to use against the service. Defaults to GET. | GET |
| serviceId | Yes | ID of the service to call (obtain from list_services). | |
| idempotencyKey | No | Optional idempotency key (UUID recommended). Identical keys within the proxy's retention window return the cached result without re-charging. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Description elaborates on behavioral traits beyond annotations: it details payment mechanics, HTTP 402 handling, idempotency via key, x402 interface, and demo mode. It confirms non-idempotent, external side effects, and spending money—fully consistent with annotations.
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 well-structured into sections (Preconditions, Behavior, x402-COMPATIBLE INTERFACE, Returns) and front-loads the core purpose. It is somewhat lengthy due to the tool's complexity, but each sentence adds value. Minor redundancy could be trimmed.
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 fully explains the return format and error states (PERMIT_MISSING, HTTP 402, payment pending). It covers all aspects: preconditions, behavior, demo mode, x402 handling, and retry logic. No gaps.
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 the schema already documents each parameter well. The description adds context like idempotencyKey retry logic and serviceId sourcing, but does not significantly enhance per-parameter understanding beyond what the schema provides.
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: 'Invoke an upstream API service through LemonCake's pay-per-call proxy.' This specific verb+resource, combined with detailed behavioral context, distinguishes it from sibling tools like check_balance, list_services, and setup.
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?
Description provides explicit preconditions (LEMON_CAKE_PERMIT env var, demo mode, serviceId from list_services) and behavior (spends money, external side effects). It indirectly guides when to use this tool vs alternatives by stating serviceId must come from list_services. Could be slightly more explicit about when not to use it, but is sufficient.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Check the current on-chain USDC balance of the wallet that owns the configured permit.
PRECONDITIONS: • DEMO MODE (no LEMON_CAKE_PERMIT): returns a canned $1.00 demo balance so trial users see something useful before signing. • LIVE: queries Base USDC.balanceOf(owner) directly + reports remaining daily permit cap. LemonCake's backend doesn't store the balance — it's read from the chain.
Use BEFORE call_service to confirm sufficient funds and remaining daily cap.
Returns: { balanceUsdc, dailyCap, remainingToday, ownerAddress, [mode], [note] }
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Beyond annotations (readOnly, idempotent), the description discloses demo mode behavior (canned $1.00) and live mode (queries chain, reports daily cap). It also notes that the backend does not store balance.
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?
Well-structured with clear sections (preconditions, usage, returns). Front-loaded with main purpose. No redundant sentences.
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?
Complete for a zero-param read-only tool. Describes all relevant behaviors, return fields, and usage context. No gaps.
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?
No parameters exist; schema coverage is 100%. The description adds value by detailing the return fields (balanceUsdc, dailyCap, etc.), compensating for the lack of output 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 specifies the exact verb 'check' and resource 'on-chain USDC balance of the wallet that owns the configured permit.' It clearly distinguishes from sibling tools like call_service or check_tax.
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 recommends using this tool BEFORE call_service to confirm funds and daily cap. Also outlines preconditions for demo vs. live mode, providing clear context for usage.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Run a Japanese tax compliance check on a single transaction. No authentication required.
Performs three checks in one call:
Validates the qualified-invoice registration number (T-number) against the NTA registry.
Determines whether source-withholding (源泉徴収) applies based on the service description.
If withholding applies, computes the withholding amount and net payable.
Intended for Japanese corporations that pay AI / API services and need to file withholding correctly under the qualified-invoice (インボイス制度) regime.
Returns: { invoice: { valid, name, ... }, withholding: { required, rate, amount, net } } Errors: invalid registrationNumber returns invoice.valid = false (not an exception).
| Name | Required | Description | Default |
|---|---|---|---|
| grossAmountJpy | Yes | Gross transaction amount in JPY, tax inclusive. | |
| registrationNumber | Yes | Qualified-invoice registration number issued by the Japanese NTA (e.g. "T1234567890123"). | |
| serviceDescription | Yes | Plain-text description of what was purchased. Used to classify whether source-withholding applies. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
The description adds significant context beyond annotations: it states no authentication required, details the three checks performed, and explains error handling (invalid registrationNumber returns invoice.valid = false, not an exception). This aligns with the readOnlyHint and idempotentHint.
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 well-structured with a bullet list of checks, clear sections, and no unnecessary words. It is concise yet informative.
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?
Despite having no output schema, the description provides a complete picture: it explains the three checks, the return object structure, error handling, and intended usage. For a complex tax compliance tool, this is highly complete.
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%, so baseline is 3. However, the description adds detailed meaning: it explains how each parameter is used in the three checks and describes the return object structure (invoice and withholding fields), which goes beyond the 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's purpose: 'Run a Japanese tax compliance check on a single transaction.' It lists three specific checks and differentiates from sibling tools, as no other tax check tool exists among the siblings.
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 states the intended use case: 'Japanese corporations that pay AI / API services' and notes that no authentication is required. It provides context for when to use it, though it could be more explicit about 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.
Return public usage statistics for every approved service on the marketplace. No authentication required.
Use this AFTER list_services and BEFORE call_service to pick a service based on real-world traction (call counts, USDC revenue, last-used timestamp).
Returns: an array of { serviceId, callCount, totalRevenueUsdc, lastCalledAt }.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare readOnlyHint=true, destructiveHint=false, openWorldHint=true, idempotentHint=true. The description adds that no authentication is required and specifies the output format, adding value beyond annotations.
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?
Three sentences: purpose, usage guidance, return format. Front-loaded with key information, no wasted words.
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?
For a tool with no parameters and no output schema, the description fully explains functionality, usage sequence, and return structure. Adequate for the complexity.
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?
No parameters, so baseline is 4. The description does not need to elaborate on parameters.
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 returns public usage statistics for every approved service. It uses specific verb 'Return' and resource 'usage statistics for every approved service', and distinguishes from siblings by providing ordering context.
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 this tool: 'Use this AFTER list_services and BEFORE call_service to pick a service based on real-world traction.' This provides clear context and exclusion of alternatives.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
List approved API services available on the LemonCake marketplace. No authentication required.
Use this BEFORE call_service to discover serviceId values and per-call USDC pricing.
When LEMON_CAKE_PERMIT is missing, 8 demo services are prepended: demo_search → Wikipedia opensearch demo_echo → httpbin.org/anything demo_fx → live FX rates (open.er-api) demo_translate → 80+ languages (MyMemory) demo_weather → current weather any lat/lon (Open-Meteo) demo_geocode → place name → lat/lon (OpenStreetMap Nominatim) demo_time → IANA timezone time + DST (worldtimeapi) demo_dictionary → English definitions / synonyms (dictionaryapi.dev) All real upstreams, no auth, free. Live users (permit set) see real marketplace entries; demo_* IDs remain callable directly.
Each item: { id, name, provider, type ('API' | 'MCP'), pricePerCall, [usage], [mode] }.
Errors: HTTP-level errors are returned as Error: API <status>: <body>.
| Name | Required | Description | Default |
|---|---|---|---|
| limit | No | Maximum number of services to return (default 50, max 100). |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already indicate read-only, idempotent, and non-destructive behavior. The description adds valuable context: no authentication required, demo services prepended when permit missing, error format, and output structure. No contradictions with annotations.
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 well-structured with a clear purpose statement, usage direction, and detailed list of demo services. It is slightly lengthy but each sentence adds value, and bullet-like formatting aids readability.
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?
For a list tool with no output schema, the description includes output structure (fields), error handling, and relationship to sibling tools. It covers the demo behavior comprehensively, making it complete for an AI agent.
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 only parameter 'limit' is fully described in the schema (type, default, min, max, description). The description does not add additional meaning beyond what the schema provides, so baseline score of 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 'List approved API services available on the LemonCake marketplace' with a specific verb and resource. It distinguishes from sibling tools like call_service by explicitly mentioning its role as a prerequisite for discovering serviceId values.
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 advises to 'Use this BEFORE call_service to discover serviceId values and per-call USDC pricing' and explains the demo service behavior when LEMON_CAKE_PERMIT is missing. It does not explicitly say when not to use it, but the context is clear.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Show the LemonCake MCP first-run setup guide. No authentication required. Call this tool FIRST to learn what is missing and how to obtain a permit.
If LEMON_CAKE_PERMIT is not set, the server is in DEMO MODE: list_services returns 8 free demo services (search / echo / fx / translate / weather / geocode / time / dictionary) powered by real upstreams (Wikipedia, httpbin, Open-Meteo, Nominatim, etc.). call_service and check_balance respond with mock data so you can verify integration before signing.
Returns the current credential status (permit presence), demo-mode flag, and step-by-step instructions for getting a permit at /start/v2, including a sample MCP client config snippet ready to paste.
Returns: { version, apiUrl, mode, credentials, availableTools, setupSteps, permitUrl, docs } Errors: none — this tool always succeeds.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Built upon annotations (readOnlyHint, idempotentHint, destructiveHint), the description adds key behaviors: no authentication required, always succeeds, returns specific fields, and explains demo mode behavior. No contradictions.
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 well-structured with a clear purpose, context on demo mode, and a list of return fields. Every sentence adds value, and it is front-loaded with the main purpose.
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 includes the complete return object fields. It covers error behavior (none), authentication needs, and the demo mode vs. full mode scenario, making it fully self-contained for a setup guide.
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?
No parameters exist, so baseline is 4. The description does not add parameter info, but that's irrelevant here. Schema coverage is 100%.
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 starts with a clear verb+resource: 'Show the LemonCake MCP first-run setup guide.' It distinguishes itself from sibling tools by stating 'Call this tool FIRST,' making its unique purpose obvious.
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 advises when to use the tool: 'Call this tool FIRST to learn what is missing and how to obtain a permit.' It explains the context of demo mode and provides guidance on what to expect based on credential status.
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.
v0.5.9call_servicecheck_balancecheck_taxget_service_statslist_servicessetupv0.5.4call_servicecheck_balancecheck_taxget_service_statslist_servicessetupv0.5.2call_servicecheck_balancecheck_taxget_service_statslist_servicessetupv0.5.1call_servicecheck_balancecheck_taxget_service_statslist_servicessetupv0.1.0call_servicecheck_balancecheck_taxget_service_statslist_servicessetupScored across 6 tools
Each tool has a clearly distinct role: setup guides onboarding, list_services discovers offerings, call_service executes paid calls, check_balance monitors funds, check_tax handles compliance, and get_service_stats provides marketplace analytics. There is no meaningful overlap or ambiguity between them.
Tool names follow a consistent snake_case verb_noun pattern: list_services, call_service, check_balance, check_tax, get_service_stats. Even 'setup' fits as a single-word imperative without breaking the overall stylistic consistency.
Six tools is a well-scoped count for a billing/proxy service: onboarding, discovery, execution, balance checking, tax compliance, and usage stats each earn their place. The server feels neither bloated nor thin.
The core lifecycle is well covered: discover services, call them, check balance, and verify tax obligations. Minor gaps exist, such as no explicit transaction-history or budget-configuration tool, but the daily cap and receipt fields mitigate these and the main workflows are complete.
30 pay-per-call APIs for AI agents: compliance, trade, safety, web, data. USDC on Base via x402.
Non-custodial stablecoin bill-pay rails on Celo & Base for AI agents, settled on-chain via MCP.
Payment infrastructure for AI agents: spending rules, approval flows, single-use virtual cards.
Wallet and payments for AI agents: auto-pay x402 APIs in USDC on XDC, within on-chain limits.
