Screwtape.ai

Présentation

Screwtape.ai est un contrôle indépendant de la sortie des agents IA. Il évalue une réponse brouillon par rapport aux preuves fournies, note la concordance entre 2 et 16 réponses pour repérer un faux consensus, et vérifie les doublons dans les factures avant qu’un agent ne les paie. Les résultats réussis peuvent inclure un reçu signé Ed25519, et un reçu peut être publié sous forme de badge de vérification.

Il fonctionne comme un endpoint Streamable HTTP hébergé à l’adresse https://screwtape.ai/mcp. Les appels utilisent une clé API Bearer (une clé gratuite autorise 100 appels par jour UTC) ou un paiement par appel de 0,005 $ en USDC sur Base ou Solana via x402. Le contenu évalué n’est pas stocké, sauf si retain est défini sur true.

Ce que vous pouvez faire

• Obtenir un verdict pass, fail ou retry sur une sortie d’agent par rapport aux preuves
• Vérifier les doublons d’une facture avant qu’un agent ne la paie
• Évaluer le faux consensus sur 2 à 16 réponses
• Créer une clé API gratuite via une inscription par preuve de travail
• Obtenir les conditions de paiement des packs d’appels prépayés
• Publier un reçu signé sous forme de badge public vérifié

Outils 9

• budget_ledgerÉcriture

  Budget Ledger

• action_receiptÉcriture

  Action Receipts

• payment_checkLecture seule

  Payment check

• judgeLecture seule

  Renvoyer un verdict pass, fail ou retry sur une sortie d’agent, vérifiée par rapport aux preuves fournies.

• check_duplicate_invoiceÉcriture

  Évaluer une facture et la comparer aux factures enregistrées précédemment pour indiquer son statut de doublon.

• consensusLecture seule

  Noter la concordance entre 2 et 16 réponses et signaler un consensus copié ou faible.

• get_api_keyÉcriture

  Créer une clé API gratuite (100 appels par jour UTC) via une preuve de travail SHA-256.

• publish_receiptÉcriture

  Publier un reçu signé dont vous êtes propriétaire sous forme de page de badge public Screwtape Verified.

• try_checkLecture seule

  Try one sample duplicate-invoice check (free, no key)

Lus directement sur le serveur le 15 nov. 2026. Lecture seule, Écriture et Peut supprimer sont des indications déclarées par le serveur ; c’est votre client qui décide de vous demander confirmation avant d’exécuter un outil.

Exemples de prompts

• Vérifie si cette facture est en double avant de la payer.

• Évalue mon brouillon de réponse par rapport à ces preuves et dis-moi s’il passe.

• Fais un contrôle de consensus sur ces cinq réponses d’agents.

• Publie le reçu du dernier verdict sous forme de badge.

Configuration

Chaque client se connecte à screwtape.ai. Choisissez le vôtre ; la page retient votre choix.

claude mcp add --transport http screwtape-ai https://screwtape.ai/mcp

https://screwtape.ai/mcp

{
  "mcpServers": {
    "screwtape-ai": {
      "url": "https://screwtape.ai/mcp"
    }
  }
}

code --add-mcp '{"name":"screwtape-ai","type":"http","url":"https://screwtape.ai/mcp"}'

{
  "servers": {
    "screwtape-ai": {
      "type": "http",
      "url": "https://screwtape.ai/mcp"
    }
  }
}

{
  "mcpServers": {
    "screwtape-ai": {
      "serverUrl": "https://screwtape.ai/mcp"
    }
  }
}

codex mcp add screwtape-ai --url https://screwtape.ai/mcp

[mcp_servers.screwtape-ai]
url = "https://screwtape.ai/mcp"

gemini mcp add --transport http screwtape-ai https://screwtape.ai/mcp

{
  "mcpServers": {
    "screwtape-ai": {
      "type": "http",
      "url": "https://screwtape.ai/mcp"
    }
  }
}

Intégrer

Affichez Screwtape.ai dans un README ou sur votre site. Chaque intégration renvoie à cette page, utilise l’URL du serveur et ne dépose aucun cookie. Comment fonctionnent les intégrations

La carte, l’iframe et le JSON sont en anglais. Ajoutez ?lang=it, fr, de ou es à leur adresse pour une autre langue.

Badge pour README Carte pour site web iframe JSON Prompt pour l’IA

Badge pour README

Markdown

[![Screwtape.ai on mcp.tc](https://mcp.tc/i/screwtape-ai/badge.svg)](https://mcp.tc/i/screwtape-ai)

HTML

<a href="https://mcp.tc/i/screwtape-ai"><img src="https://mcp.tc/i/screwtape-ai/badge.svg" alt="Screwtape.ai on mcp.tc" height="20"></a>

Il suit le mode clair ou sombre du lecteur. Ajoutez ?theme=light ou ?theme=dark pour le fixer, ou ?style=compact pour afficher le symbole sans le nom.

Carte pour site web

HTML

<script src="https://mcp.tc/w/screwtape-ai.js" async></script>

La carte apparaît à l’endroit de la balise. Ajoutez data-theme="dark" ou data-size="compact" à la balise pour la modifier. Le fichier contient tout ce qu’il lui faut : il ne fait aucune autre requête et ne dépose aucun cookie.

iframe

HTML

<iframe src="https://mcp.tc/embed/screwtape-ai" title="Screwtape.ai on mcp.tc" width="420" height="200" loading="lazy" allow="clipboard-write" style="border:0;border-radius:8px;max-width:100%"></iframe>

Utilisez-la sur les pages qui n’autorisent pas les scripts. Ajoutez ?theme=light ou ?theme=dark à l’adresse pour fixer les couleurs.

JSON

La fiche sous forme de données pour vos pages et vos outils : nom, accroche, URL du serveur, outils et liens « Add to ». N’importe quel site peut la lire (CORS ouvert).

https://mcp.tc/i/screwtape-ai.json

Ce que signifie chaque champ

Prompt pour l’IA

Collez-le dans un assistant de code IA, comme Claude Code ou Cursor, et il ajoutera la carte de Screwtape.ai à votre site.

Markdown

Ajoute à mon site la carte mcp.tc de Screwtape.ai, un serveur MCP. C’est un seul script statique : pas de build, aucun paquet à installer, aucun cookie.

Place cette balise là où la carte doit apparaître. La carte s’affiche juste avant elle :

```html
<script src="https://mcp.tc/w/screwtape-ai.js" async></script>
```

Pour l’afficher ailleurs, mets ce conteneur à cet endroit et charge le même script une seule fois, par exemple avant `</body>`. Garde le lien à l’intérieur : les visiteurs sans JavaScript le voient.

```html
<div data-mcptc="screwtape-ai"><a href="https://mcp.tc/i/screwtape-ai">Screwtape.ai on mcp.tc</a></div>
```

Options :
- `data-theme="light"` ou `"dark"` sur la balise ou le conteneur fixe les couleurs. Sans cela, la carte suit le mode clair ou sombre du visiteur.
- `data-size="compact"` donne une carte plus petite, sans l’accroche.
- La variable CSS `--mcptc-max-width` sur le conteneur ou un parent fixe la largeur (`520px` par défaut).
- `?lang=` dans l’URL du script choisit la langue de la carte : `en`, `it`, `fr`, `de` ou `es`.

Si le site a une Content Security Policy, autorise `script-src https://mcp.tc` et `img-src data:` (l’icône est intégrée).

Pour un README ou une autre doc en Markdown, utilise plutôt le badge :

```markdown
[![Screwtape.ai on mcp.tc](https://mcp.tc/i/screwtape-ai/badge.svg)](https://mcp.tc/i/screwtape-ai)
```

Référence : https://mcp.tc/fr/embed

Historique des changements

Nous lisons la liste d’outils de ce serveur toutes les 3 heures et notons ce qui change. Dernière lecture : 15 nov. 2026, 00:16 UTC.

1. 10 nov. 2026, 05:48 UTC, Vérification automatique

   1 outil ajouté. Instructions du serveur modifiées.

   Nouvelles instructions du serveur

   Détails, 10 nov. 2026, 05:48 UTC

   Ajoutés 1

   • payment_checkLecture seule

     Payment check: is this seller safe to pay, checked at the moment the agent pays. Compares the 402 price and payee with the seller's published listings when they can be fetched, screens the payee against a local denylist and the bundled OFAC SDN digital-currency address list, flags an over-price versus the intended amount, reuses the duplicate-payment check, and confirms the seller URL returns a valid 402. Returns allow, review or block with reasons and a signed receipt id. Fetches time out in about 2s and fail closed to review. $0.005 per call.

   Instructions du serveur

   Avant

   Screwtape.ai decides whether an AI agent is allowed to spend money, and keeps the proof. Call check_duplicate_invoice (Bearer key or x402 payment) before an agent pays an invoice or payment instruction: it returns duplicate.status (unique, duplicate, possible_duplicate, replay), runs the checks you pass (for example a sum check that line items add up to the total) and judges the invoice against your evidence, such as the purchase order. Pay only when next_step.action is pass; on escalate, do not pay. Every verdict carries a signed receipt as proof of why the money moved or did not. Use budget_ledger to reserve cumulative spend ($0.002); action_receipt signs an action record ($0.003). All five paid tools share 3 free calls per UTC day without a key plus 10/day with a free key. Anonymous: private daily IP namespace, 3/minute, global 2000/day. Free GET /v1/receipt/verify checks action receipts and /.well-known/screwtape-receipt-key publishes their key. Use judge on one draft answer before you return it (send evidence; it does not search the web). Use consensus when 2 to 16 answers to the same task agree and you are about to trust the majority. Use get_api_key only if no key is configured and you can compute SHA-256 in code. judge, check_duplicate_invoice and consensus cost one call each: a free key gives 10 per UTC day, or pay $0.005 USDC on Base or Solana with x402. After anonymous access is exhausted, an unpaid tools/call returns HTTP 200 with isError true and the x402 PaymentRequirements in structuredContent and result._meta['x402/payment-required']. Retry with params._meta['x402/payment']. Settlement is echoed in result._meta['x402/payment-response']. Judged content is not stored unless retain is true or the key owner turned on opt-in data sharing (off by default; redacted, kept 90 days). A successful judge or consensus result includes an Ed25519 receipt when signing is configured. Public keys are at https://screwtape.ai/.well-known/screwtape-receipt-keys.json. PO…

   Après

   Screwtape.ai decides whether an AI agent is allowed to spend money, and keeps the proof. Call check_duplicate_invoice (Bearer key or x402 payment) before an agent pays an invoice or payment instruction: it returns duplicate.status (unique, duplicate, possible_duplicate, replay), runs the checks you pass (for example a sum check that line items add up to the total) and judges the invoice against your evidence, such as the purchase order. Pay only when next_step.action is pass; on escalate, do not pay. Every verdict carries a signed receipt as proof of why the money moved or did not. Use budget_ledger to reserve cumulative spend ($0.002); action_receipt signs an action record ($0.003). Use payment_check ($0.005) before paying a seller, to check the 402 payee and price. All six paid tools share 3 free calls per UTC day without a key plus 10/day with a free key. Anonymous: private daily IP namespace, 3/minute, global 2000/day. Free GET /v1/receipt/verify checks action receipts and /.well-known/screwtape-receipt-key publishes their key. Use judge on one draft answer before you return it (send evidence; it does not search the web). Use consensus when 2 to 16 answers to the same task agree and you are about to trust the majority. Use get_api_key only if no key is configured and you can compute SHA-256 in code. judge, check_duplicate_invoice and consensus cost one call each: a free key gives 10 per UTC day, or pay $0.005 USDC on Base or Solana with x402. After anonymous access is exhausted, an unpaid tools/call returns HTTP 200 with isError true and the x402 PaymentRequirements in structuredContent and result._meta['x402/payment-required']. Retry with params._meta['x402/payment']. Settlement is echoed in result._meta['x402/payment-response']. Judged content is not stored unless retain is true or the key owner turned on opt-in data sharing (off by default; redacted, kept 90 days). A successful judge or consensus result includes an Ed25519 receipt when signing is configured.…

2. 9 nov. 2026, 14:50 UTC, Vérification automatique

   Nouvelle version 1.3.0 (avant : 1.2.2). Annoncée par le serveur.

3. 9 nov. 2026, 14:50 UTC, Vérification automatique

   1 outil modifié.

   Détails, 9 nov. 2026, 14:50 UTC

   Modifiés 1

   • action_receiptdescription, schéma d’entrée et schéma de sortie
     Avant

     Action Receipts: Ed25519-signed, hash-chained attestation of an agent-reported action and status; not evidence the action happened or a legal compliance certification. Per caller and agent chain; concurrent appends serialize with compare-and-set. Raw action/reference are hashed, not stored. event_id per agent is the authoritative idempotency key; a custom HTTP key cannot override it. Same event_id/body returns the same receipt; changed body conflicts. GET /v1/receipt/verify?id=<receipt_id> is free and returns the receipt to anyone who knows its unguessable ID. Or pass receipt=<JSON> to verify a portable receipt. Keep IDs private if your agent IDs/status are sensitive. Records persist until the API key is deleted; keyless records have no automatic expiry. $0.003 per call; shared 3 free calls per UTC day without a key (daily private IP namespace), plus 10/day with a free key. Anonymous scopes rotate daily; use a key for a persistent ledger/chain. Send the same Bearer key plus a payment…

     Après

     Action Receipts: Ed25519-signed, hash-chained attestation of an agent-reported action and status; not evidence the action happened or a legal compliance certification. Per caller and agent chain; concurrent appends serialize with compare-and-set. Raw action/reference are hashed, not stored. event_id per agent is the authoritative idempotency key; a custom HTTP key cannot override it. Same event_id/body returns the same receipt; changed body conflicts. Optional children (1-64 receipt UUIDs) closes a run: the signed children_root digests ordered child receipt_hash values so any added, dropped or swapped step fails verification. GET /v1/receipt/verify?id=<receipt_id> is free and returns the receipt to anyone who knows its unguessable ID; children_root_valid is set when children are present. A receipt proves the record is unchanged since issued, not that a payment was correct or an outside action happened. Or pass receipt=<JSON> to verify a portable receipt. Keep IDs private if your agent…

     • Nouveaux paramètres : children

4. 7 nov. 2026, 14:47 UTC, Vérification automatique

   Nouvelle version 1.2.2 (avant : 1.2.0). Annoncée par le serveur.

5. 7 nov. 2026, 02:45 UTC, Vérification automatique

   2 outils ajoutés et 3 outils modifiés. Instructions du serveur modifiées.

   Nouvelles instructions du serveur

   Détails, 7 nov. 2026, 02:45 UTC

   Ajoutés 2

   • action_receiptÉcriture

     Action Receipts: Ed25519-signed, hash-chained attestation of an agent-reported action and status; not evidence the action happened or a legal compliance certification. Per caller and agent chain; concurrent appends serialize with compare-and-set. Raw action/reference are hashed, not stored. event_id per agent is the authoritative idempotency key; a custom HTTP key cannot override it. Same event_id/body returns the same receipt; changed body conflicts. GET /v1/receipt/verify?id=<receipt_id> is free and returns the receipt to anyone who knows its unguessable ID. Or pass receipt=<JSON> to verify a portable receipt. Keep IDs private if your agent IDs/status are sensitive. Records persist until the API key is deleted; keyless records have no automatic expiry. $0.003 per call; shared 3 free calls per UTC day without a key (daily private IP namespace), plus 10/day with a free key. Anonymous scopes rotate daily; use a key for a persistent ledger/chain. Send the same Bearer key plus a payment…

   • budget_ledgerÉcriture

     Budget Ledger: atomically reserves reported USD spend across a run and its sub-agents. Returns allow/warn/block, remaining budget, retry storms and spend velocity. Call judge first, then reserve immediately before spending; honor block. This is cooperative accounting, not payment enforcement. Root cap and parent are immutable. Children inherit the root cap (send the same cap). event_id is the authoritative idempotency key; a custom HTTP key cannot override it. New event_id per attempted action; same event_id and body replays without reserving twice. No refunds or cap resets. Warn reservations count as spend. Limits: USD, six decimals, cap at most 1000000. Three matching actions in 60 seconds allowed; fourth blocked. Velocity warning at ten attempts/minute or more than 25% of cap/minute; 80% usage warns. Use action to distinguish operations at a merchant. Stored: IDs, keyed fingerprints and amounts, not raw action/merchant. Records persist until the API key is deleted; keyless records…

   Modifiés 3

   • check_duplicate_invoicedescription
     Avant

     Spend gate for invoices: decides whether an agent may pay this invoice, with a signed receipt as proof. Same as judge with one duplicate check: it compares this invoice with the invoices this caller recorded earlier and also judges it. Call it before an agent pays or approves an invoice or payment instruction. Inputs: the judge inputs (task, agent_output as the invoice JSON, evidence, criteria, checks, mode, tags) plus duplicate {vendor_path, invoice_id_path, amount_path, currency_path?, date_path?, payee_account_path?, accept_payee_change?, window_days?, store?, retain_days?}. Returns the judge result plus duplicate.status (unique, duplicate, possible_duplicate, replay or not_checked). On duplicate or possible_duplicate, next_step.action is escalate: do not pay. Fingerprints are kept per caller; only keyed HMAC fingerprints are stored, and only when store is true. Cost: 1 call, the same as judge. Needs Authorization: Bearer <key>, or a per-call payment in params._meta with no key (re…

     Après

     Spend gate for invoices: decides whether an agent may pay this invoice, with a signed receipt as proof. Same as judge with one duplicate check: it compares this invoice with the invoices this caller recorded earlier and also judges it. Call it before an agent pays or approves an invoice or payment instruction. Inputs: the judge inputs (task, agent_output as the invoice JSON, evidence, criteria, checks, mode, tags) plus duplicate {vendor_path, invoice_id_path, amount_path, currency_path?, date_path?, payee_account_path?, accept_payee_change?, window_days?, store?, retain_days?}. Returns the judge result plus duplicate.status (unique, duplicate, possible_duplicate, replay or not_checked). On duplicate or possible_duplicate, next_step.action is escalate: do not pay. Fingerprints are kept per caller; only keyed HMAC fingerprints are stored, and only when store is true. Cost: 1 call, the same as judge. First 3 calls per UTC day need no key (daily private IP namespace). Then use Authorizati…

   • consensusdescription

     Le texte affiché ici est identique avant et après : le changement se situe au-delà des 1 000 premiers caractères ou dans des caractères qui ne s’affichent pas.

   • judgedescription
     Avant

     A pass/fail/retry verdict on ONE agent output, from a service outside your agent, used as the go/no-go before an agent spends money or acts on that output. Call it before you pay on, return, store or act on a draft answer, especially when you have the evidence the answer should be based on. Do NOT call it to look up facts (it does no web search and only checks against the evidence you pass), to compare several agents' answers (use consensus), or for offline evaluation over a dataset. Inputs: task, agent_output, evidence, criteria, retain, and optional agent_id and version tags. Returns: verdict, score, reasons, engine, next_step, request_id, retention, the tags when sent, and receipt when signing is configured. On pass, use the output. On retry, revise using reasons and call again, at most 2 or 3 times. On fail, stop. Optional checks run first; on failure you get next_step.hint saying what to fix. Branch on next_step.action (pass, retry or escalate). Cost: 1 call. Needs Authorization:…

     Après

     A pass/fail/retry verdict on ONE agent output, from a service outside your agent, used as the go/no-go before an agent spends money or acts on that output. Call it before you pay on, return, store or act on a draft answer, especially when you have the evidence the answer should be based on. Do NOT call it to look up facts (it does no web search and only checks against the evidence you pass), to compare several agents' answers (use consensus), or for offline evaluation over a dataset. Inputs: task, agent_output, evidence, criteria, retain, and optional agent_id and version tags. Returns: verdict, score, reasons, engine, next_step, request_id, retention, the tags when sent, and receipt when signing is configured. On pass, use the output. On retry, revise using reasons and call again, at most 2 or 3 times. On fail, stop. Optional checks run first; on failure you get next_step.hint saying what to fix. Branch on next_step.action (pass, retry or escalate). Cost: 1 call. First 3 calls per UT…

   Instructions du serveur

   Avant

   Screwtape.ai decides whether an AI agent is allowed to spend money, and keeps the proof. Call check_duplicate_invoice (Bearer key or x402 payment) before an agent pays an invoice or payment instruction: it returns duplicate.status (unique, duplicate, possible_duplicate, replay), runs the checks you pass (for example a sum check that line items add up to the total) and judges the invoice against your evidence, such as the purchase order. Pay only when next_step.action is pass; on escalate, do not pay. Every verdict carries a signed receipt as proof of why the money moved or did not. Use judge on one draft answer before you return it (send evidence; it does not search the web). Use consensus when 2 to 16 answers to the same task agree and you are about to trust the majority. Use get_api_key only if no key is configured and you can compute SHA-256 in code. judge, check_duplicate_invoice and consensus cost one call each: a free key gives 10 per UTC day, or pay $0.005 USDC on Base or Solana with x402. An unpaid tools/call returns HTTP 200 with isError true and the x402 PaymentRequirements in structuredContent and result._meta['x402/payment-required']. Retry with params._meta['x402/payment']. Settlement is echoed in result._meta['x402/payment-response']. Judged content is not stored unless retain is true or the key owner turned on opt-in data sharing (off by default; redacted, kept 90 days). A successful judge or consensus result includes an Ed25519 receipt when signing is configured. Public keys are at https://screwtape.ai/.well-known/screwtape-receipt-keys.json. POST /v1/receipts/verify checks a receipt. publish_receipt publishes one receipt you own. It needs the Bearer key and does not accept x402.

   Après

   Screwtape.ai decides whether an AI agent is allowed to spend money, and keeps the proof. Call check_duplicate_invoice (Bearer key or x402 payment) before an agent pays an invoice or payment instruction: it returns duplicate.status (unique, duplicate, possible_duplicate, replay), runs the checks you pass (for example a sum check that line items add up to the total) and judges the invoice against your evidence, such as the purchase order. Pay only when next_step.action is pass; on escalate, do not pay. Every verdict carries a signed receipt as proof of why the money moved or did not. Use budget_ledger to reserve cumulative spend ($0.002); action_receipt signs an action record ($0.003). All five paid tools share 3 free calls per UTC day without a key plus 10/day with a free key. Anonymous: private daily IP namespace, 3/minute, global 2000/day. Free GET /v1/receipt/verify checks action receipts and /.well-known/screwtape-receipt-key publishes their key. Use judge on one draft answer before you return it (send evidence; it does not search the web). Use consensus when 2 to 16 answers to the same task agree and you are about to trust the majority. Use get_api_key only if no key is configured and you can compute SHA-256 in code. judge, check_duplicate_invoice and consensus cost one call each: a free key gives 10 per UTC day, or pay $0.005 USDC on Base or Solana with x402. After anonymous access is exhausted, an unpaid tools/call returns HTTP 200 with isError true and the x402 PaymentRequirements in structuredContent and result._meta['x402/payment-required']. Retry with params._meta['x402/payment']. Settlement is echoed in result._meta['x402/payment-response']. Judged content is not stored unless retain is true or the key owner turned on opt-in data sharing (off by default; redacted, kept 90 days). A successful judge or consensus result includes an Ed25519 receipt when signing is configured. Public keys are at https://screwtape.ai/.well-known/screwtape-receipt-keys.json. PO…

6. 6 nov. 2026, 20:45 UTC, Vérification automatique

   Instructions du serveur modifiées.

   Nouvelles instructions du serveur

   Détails, 6 nov. 2026, 20:45 UTC

   Instructions du serveur

   Avant

   Screwtape.ai decides whether an AI agent is allowed to spend money, and keeps the proof. Call check_duplicate_invoice (Bearer key or x402 payment) before an agent pays an invoice or payment instruction: it returns duplicate.status (unique, duplicate, possible_duplicate, replay), runs the checks you pass (for example a sum check that line items add up to the total) and judges the invoice against your evidence, such as the purchase order. Pay only when next_step.action is pass; on escalate, do not pay. Every verdict carries a signed receipt as proof of why the money moved or did not. Use judge on one draft answer before you return it (send evidence; it does not search the web). Use consensus when 2 to 16 answers to the same task agree and you are about to trust the majority. Use get_api_key only if no key is configured and you can compute SHA-256 in code. judge, check_duplicate_invoice and consensus cost one call each: a free key gives 10 per UTC day, or pay $0.005 USDC on Base or Solana with x402. An unpaid tools/call returns HTTP 200 with isError true and the x402 PaymentRequirements in structuredContent. Retry with params._meta['x402/payment']. Settlement is echoed in result._meta['x402/payment-response']. Judged content is not stored unless retain is true or the key owner turned on opt-in data sharing (off by default; redacted, kept 90 days). A successful judge or consensus result includes an Ed25519 receipt when signing is configured. Public keys are at https://screwtape.ai/.well-known/screwtape-receipt-keys.json. POST /v1/receipts/verify checks a receipt. publish_receipt publishes one receipt you own. It needs the Bearer key and does not accept x402.

   Après

   Screwtape.ai decides whether an AI agent is allowed to spend money, and keeps the proof. Call check_duplicate_invoice (Bearer key or x402 payment) before an agent pays an invoice or payment instruction: it returns duplicate.status (unique, duplicate, possible_duplicate, replay), runs the checks you pass (for example a sum check that line items add up to the total) and judges the invoice against your evidence, such as the purchase order. Pay only when next_step.action is pass; on escalate, do not pay. Every verdict carries a signed receipt as proof of why the money moved or did not. Use judge on one draft answer before you return it (send evidence; it does not search the web). Use consensus when 2 to 16 answers to the same task agree and you are about to trust the majority. Use get_api_key only if no key is configured and you can compute SHA-256 in code. judge, check_duplicate_invoice and consensus cost one call each: a free key gives 10 per UTC day, or pay $0.005 USDC on Base or Solana with x402. An unpaid tools/call returns HTTP 200 with isError true and the x402 PaymentRequirements in structuredContent and result._meta['x402/payment-required']. Retry with params._meta['x402/payment']. Settlement is echoed in result._meta['x402/payment-response']. Judged content is not stored unless retain is true or the key owner turned on opt-in data sharing (off by default; redacted, kept 90 days). A successful judge or consensus result includes an Ed25519 receipt when signing is configured. Public keys are at https://screwtape.ai/.well-known/screwtape-receipt-keys.json. POST /v1/receipts/verify checks a receipt. publish_receipt publishes one receipt you own. It needs the Bearer key and does not accept x402.

7. 5 nov. 2026, 23:46 UTC, Vérification automatique

   1 outil ajouté, 1 outil retiré et 4 outils modifiés. Instructions du serveur modifiées.

   Nouvelles instructions du serveur

   Détails, 5 nov. 2026, 23:46 UTC

   Ajoutés 1

   • try_checkLecture seule

     See what a spend check returns before you have a key. Runs ONE sample check on a fixed demo invoice: a duplicate match against an invoice paid earlier, required fields, and a sum check that the line items add up to the total. Call it once to learn the response shape. Do NOT call it on your own invoices (it ignores them; use check_duplicate_invoice) or in a loop: it is limited to 5 calls per hour per IP and per MCP session. Inputs: scenario (optional: duplicate, the default, clean, or bad_total). Returns: demo, scenario, invoice, earlier_invoice, duplicate.status (unique or duplicate), checks, next_step.action (pass, retry or escalate), note, next_action and request_id. No receipt and nothing is stored. Cost: free. Needs no key.

   Retirés 1

   • buy_pack

     Retired. No prepaid pack is sold to agents any more (Starter and Builder were retired), so this tool always returns the error pack_retired with how a person buys the Sandbox Evaluation Pack by card. It does not move funds and needs no key. Kept so existing clients get a clear answer. To pay per call without a key, call judge or consensus and handle the x402 payment requirements.

   Modifiés 4

   • check_duplicate_invoicedescription, schéma d’entrée et schéma de sortie
     Avant

     Duplicate invoice check before an agent pays. Same as judge with one duplicate check: it compares this invoice with the invoices this API key recorded earlier and also judges it. Call it before an agent pays or approves an invoice or payment instruction. Inputs: the judge inputs (task, agent_output as the invoice JSON, evidence, criteria, checks, mode, tags) plus duplicate {vendor_path, invoice_id_path, amount_path, currency_path?, date_path?, window_days?, store?, retain_days?}. Returns the judge result plus duplicate.status (unique, duplicate, possible_duplicate, replay or not_checked). On duplicate or possible_duplicate, next_step.action is escalate: do not pay. Needs Authorization: Bearer, because fingerprints are kept per key; only keyed HMAC fingerprints are stored, and only when store is true. Cost: 1 call, the same price as judge. Without a key or payment, tools/call is HTTP 200 with isError true and the x402 PaymentRequirements in structuredContent; an x402 payment without a…

     Après

     Spend gate for invoices: decides whether an agent may pay this invoice, with a signed receipt as proof. Same as judge with one duplicate check: it compares this invoice with the invoices this caller recorded earlier and also judges it. Call it before an agent pays or approves an invoice or payment instruction. Inputs: the judge inputs (task, agent_output as the invoice JSON, evidence, criteria, checks, mode, tags) plus duplicate {vendor_path, invoice_id_path, amount_path, currency_path?, date_path?, payee_account_path?, accept_payee_change?, window_days?, store?, retain_days?}. Returns the judge result plus duplicate.status (unique, duplicate, possible_duplicate, replay or not_checked). On duplicate or possible_duplicate, next_step.action is escalate: do not pay. Fingerprints are kept per caller; only keyed HMAC fingerprints are stored, and only when store is true. Cost: 1 call, the same as judge. Needs Authorization: Bearer <key>, or a per-call payment in params._meta with no key (re…

     • Description de checks
       Avant

       Optional. The worker runs each check. A failed check forces verdict fail. Types: url_contains, json_schema, equals, regex, numeric, onchain, required_fields, types, format, sum, duplicate. duplicate…

       Après

       Optional. The worker runs each check. A failed check forces verdict fail. Types include url_contains, json_schema, equals, regex, numeric, required_fields, types, format, sum and duplicate. duplicate…

     • Description de duplicate
       Avant

       The duplicate check: dot paths into agent_output for vendor, invoice id and amount, plus optional currency_path, date_path, window_days, store and retain_days. store true records this invoice's keyed…

       Après

       The duplicate check: dot paths into agent_output for vendor, invoice id and amount, plus optional currency_path, date_path, payee_account_path, accept_payee_change, window_days, store and retain_days…

   • consensusdescription, schéma d’entrée et schéma de sortie
     Avant

     False-consensus check for 2 to 16 answers to the same task. It scores how the answers agree and flags agreement that looks copied or too thin to trust. Screwtape is not tied to the agent framework or the model that wrote the answers. Call it when several agents, samples or tools answered the same question and you are about to accept the majority. Do NOT call it with a single answer (use judge), with answers to different tasks, or as proof that an answer is correct: high agreement is not evidence. Inputs: task, answers (2 to 16 objects {agent_id?: string, output: string}), optional evidence (a string or an array of strings, used when scoring), retain, and optional call-level agent_id and version tags. The call-level agent_id is not the per-answer agent_id. Returns: consensus_answer, agreement_score, false_consensus_risk (low, medium or high), reasons, clusters, engine, request_id, retention, the tags when sent, and receipt when signing is configured. In that receipt, verdict is false_c…

     Après

     False-consensus check for 2 to 16 answers to the same task, so a group of agents does not approve a spend just because they copied each other. It scores how the answers agree and flags agreement that looks copied or too thin to trust. Screwtape is not tied to the agent framework or the model that wrote the answers. Call it when several agents, samples or tools answered the same question and you are about to accept the majority. Do NOT call it with a single answer (use judge), with answers to different tasks, or as proof that an answer is correct: high agreement is not evidence. Inputs: task, answers (2 to 16 objects {agent_id?: string, output: string}), optional evidence (a string or an array of strings, used when scoring), retain, and optional call-level agent_id and version tags. The call-level agent_id is not the per-answer agent_id. Returns: consensus_answer, agreement_score, false_consensus_risk (low, medium or high), reasons, clusters, engine, request_id, retention, the tags whe…

     • Description de answers
       Avant

       2 to 16 answers. Each is {agent_id (optional, string), output (string)}. Use one entry per independent agent or sample.

       Après

       2 to 16 answers. Each is {agent_id (optional, string), output (string)}. Use one entry per agent or sample.

   • judgedescription, schéma d’entrée et schéma de sortie
     Avant

     An independent pass/fail/retry verdict on ONE agent output. Call it before you return, store or act on a draft answer, especially when you have the evidence the answer should be based on. Do NOT call it to look up facts (it does no web search and only checks against the evidence you pass), to compare several agents' answers (use consensus), or for offline evaluation over a dataset. Inputs: task, agent_output, evidence, criteria, retain, and optional agent_id and version tags. Returns: verdict, score, reasons, engine, next_step, request_id, retention, the tags when sent, and receipt when signing is configured. On pass, use the output. On retry, revise using reasons and call again, at most 2 or 3 times. On fail, stop. Optional checks run first; on failure you get next_step.hint saying what to fix. Branch on next_step.action (pass, retry or escalate). Cost: 1 call. A free key covers 10 calls per UTC day; without a key, pay $0.005 USDC on Base or Solana per call with x402. Without a key o…

     Après

     A pass/fail/retry verdict on ONE agent output, from a service outside your agent, used as the go/no-go before an agent spends money or acts on that output. Call it before you pay on, return, store or act on a draft answer, especially when you have the evidence the answer should be based on. Do NOT call it to look up facts (it does no web search and only checks against the evidence you pass), to compare several agents' answers (use consensus), or for offline evaluation over a dataset. Inputs: task, agent_output, evidence, criteria, retain, and optional agent_id and version tags. Returns: verdict, score, reasons, engine, next_step, request_id, retention, the tags when sent, and receipt when signing is configured. On pass, use the output. On retry, revise using reasons and call again, at most 2 or 3 times. On fail, stop. Optional checks run first; on failure you get next_step.hint saying what to fix. Branch on next_step.action (pass, retry or escalate). Cost: 1 call. Needs Authorization:…

     • Description de checks
       Avant

       Optional. The worker runs each check. A failed check forces verdict fail. Types: url_contains, json_schema, equals, regex, numeric, onchain, required_fields, types, format, sum, duplicate. duplicate…

       Après

       Optional. The worker runs each check. A failed check forces verdict fail. Types include url_contains, json_schema, equals, regex, numeric, required_fields, types, format, sum and duplicate. duplicate…

   • publish_receiptdescription
     Avant

     Publish a signed receipt you own so a public page and a Screwtape Verified badge can link to that one verdict. Call it when you want to show that a specific output received a specific verdict. Do NOT call it to certify an agent in general: the page says this output received this verdict at this time, and it is not a general safety certification of the agent. Nothing is public until this call. Inputs: receipt_id (required), include_content (optional boolean, default false; true stores and shows the task and agent output only if that call was retained), label (optional, at most 80 characters), unpublish (optional boolean; true removes the public page). Returns public_url, badge_svg_url, embed_html, and embed_markdown. Requires the API key that owns the receipt. It does not accept x402 payment. Cost: free.

     Après

     Publish a signed receipt you own so a public page and a Screwtape Verified badge can link to that one verdict. Call it when you want to show that a specific output received a specific verdict. Do NOT call it to certify an agent in general: the page says this output received this verdict at this time, and it is not a general safety certification of the agent. Nothing is public until this call. Inputs: receipt_id (required), include_content (optional boolean, default false; true stores and shows the task and agent output only if that call was retained), label (optional, at most 80 characters), unpublish (optional boolean; true removes the public page). Returns public_url, badge_svg_url, embed_html, and embed_markdown. Requires the API key that owns the receipt. Cost: free.

   Instructions du serveur

   Avant

   Verify before the agent pays: a separate check on agent work before a payment goes out (invoices, duplicate invoices, reconciliation), with signed receipts. Use check_duplicate_invoice with a Bearer key before an agent pays an invoice; it returns duplicate.status and costs one call like judge. Screwtape also gives false-consensus checks and pass/fail/retry verdicts for any agent output. Use judge on one draft answer before you return it (send evidence; it does not search the web). Use consensus when 2 to 16 answers to the same task agree and you are about to trust the majority. Use get_api_key only if no key is configured and you can compute SHA-256 in code. judge and consensus cost one call each: a free key gives 100 per UTC day, or pay $0.005 USDC on Base or Solana with x402. An unpaid tools/call returns HTTP 200 with isError true and the x402 PaymentRequirements in structuredContent. Retry with params._meta['x402/payment']. Settlement is echoed in result._meta['x402/payment-response']. Judged content is not stored unless retain is true or the key owner turned on opt-in data sharing (off by default; redacted, kept 90 days). A successful judge or consensus result includes an Ed25519 receipt when signing is configured. Public keys are at https://screwtape.ai/.well-known/screwtape-receipt-keys.json. POST /v1/receipts/verify checks a receipt. publish_receipt publishes one receipt you own. It needs the Bearer key and does not accept x402. Not the arXiv paper named Screwtape.

   Après

   Screwtape.ai decides whether an AI agent is allowed to spend money, and keeps the proof. Call check_duplicate_invoice (Bearer key or x402 payment) before an agent pays an invoice or payment instruction: it returns duplicate.status (unique, duplicate, possible_duplicate, replay), runs the checks you pass (for example a sum check that line items add up to the total) and judges the invoice against your evidence, such as the purchase order. Pay only when next_step.action is pass; on escalate, do not pay. Every verdict carries a signed receipt as proof of why the money moved or did not. Use judge on one draft answer before you return it (send evidence; it does not search the web). Use consensus when 2 to 16 answers to the same task agree and you are about to trust the majority. Use get_api_key only if no key is configured and you can compute SHA-256 in code. judge, check_duplicate_invoice and consensus cost one call each: a free key gives 10 per UTC day, or pay $0.005 USDC on Base or Solana with x402. An unpaid tools/call returns HTTP 200 with isError true and the x402 PaymentRequirements in structuredContent. Retry with params._meta['x402/payment']. Settlement is echoed in result._meta['x402/payment-response']. Judged content is not stored unless retain is true or the key owner turned on opt-in data sharing (off by default; redacted, kept 90 days). A successful judge or consensus result includes an Ed25519 receipt when signing is configured. Public keys are at https://screwtape.ai/.well-known/screwtape-receipt-keys.json. POST /v1/receipts/verify checks a receipt. publish_receipt publishes one receipt you own. It needs the Bearer key and does not accept x402.

8. 27 oct. 2026, 11:59 UTC, Vérification automatique

   1 outil modifié.

   Détails, 27 oct. 2026, 11:59 UTC

   Modifiés 1

   • buy_packdescription
     Avant

     Retired. No prepaid pack is sold to agents any more (Starter and Builder were retired), so this tool always returns the error pack_retired with how a person buys the Evaluation Pack by card. It does not move funds and needs no key. Kept so existing clients get a clear answer. To pay per call without a key, call judge or consensus and handle the x402 payment requirements.

     Après

     Retired. No prepaid pack is sold to agents any more (Starter and Builder were retired), so this tool always returns the error pack_retired with how a person buys the Sandbox Evaluation Pack by card. It does not move funds and needs no key. Kept so existing clients get a clear answer. To pay per call without a key, call judge or consensus and handle the x402 payment requirements.

9. 27 oct. 2026, 05:41 UTC, Vérification automatique

   1 outil modifié.

   Détails, 27 oct. 2026, 05:41 UTC

   Modifiés 1

   • buy_packdescription, titre, schéma d’entrée et indications
     Avant

     Return the x402 payment requirements for the Starter prepaid pack ($40, 10000 calls), plus how to pay with USDC on Base or Solana. pack is starter. This call does not move funds and does not require a key or a payment. POST /v1/packs/{pack}/buy with PAYMENT-SIGNATURE to pay. Pay with PayPal at POST /v1/packs/checkout.

     Après

     Retired. No prepaid pack is sold to agents any more (Starter and Builder were retired), so this tool always returns the error pack_retired with how a person buys the Evaluation Pack by card. It does not move funds and needs no key. Kept so existing clients get a clear answer. To pay per call without a key, call judge or consensus and handle the x402 payment requirements.

     titre

     Avant

     Buy a prepaid pack with USDC

     Après

     Retired: prepaid packs are no longer sold to agents

     • Plus obligatoires : pack
     • Description de pack
       Avant

       starter ($40, 10000 calls).

       Après

       Any former pack id. Ignored: every pack answers pack_retired.

10. 26 oct. 2026, 05:44 UTC, Vérification automatique

    4 outils modifiés.

    Détails, 26 oct. 2026, 05:44 UTC

    Modifiés 4

    • buy_packdescription et schéma d’entrée
      Avant

      Return the x402 payment requirements for a Starter ($40, 10000 calls) or Builder ($175, 50000 calls) prepaid pack, plus how to pay with USDC on Base or Solana. pack is starter or builder. This call does not move funds and does not require a key or a payment. POST /v1/packs/{pack}/buy with PAYMENT-SIGNATURE to pay. Pay with PayPal at POST /v1/packs/checkout.

      Après

      Return the x402 payment requirements for the Starter prepaid pack ($40, 10000 calls), plus how to pay with USDC on Base or Solana. pack is starter. This call does not move funds and does not require a key or a payment. POST /v1/packs/{pack}/buy with PAYMENT-SIGNATURE to pay. Pay with PayPal at POST /v1/packs/checkout.

      • Description de pack
        Avant

        starter ($40, 10000 calls) or builder ($175, 50000 calls).

        Après

        starter ($40, 10000 calls).

    • consensusdescription

      Le texte affiché ici est identique avant et après : le changement se situe au-delà des 1 000 premiers caractères ou dans des caractères qui ne s’affichent pas.

    • get_api_keydescription
      Avant

      Create a free Screwtape.ai API key (100 calls per UTC day) with no human approval. Call it only when no Screwtape key is configured and you can compute SHA-256 with a code tool, because the proof of work must be computed exactly, not guessed. Do NOT call it when a key already exists, to get more quota (it is limited to 5 keys per IP per hour and per UTC day), or without a valid pow_nonce. A missing nonce returns an error that describes the challenge. Inputs: pow_nonce (required string matching ^[A-Za-z0-9_-]{8,64}$; SHA-256 of the UTF-8 string screwtape-signup-v1:<current UTC YYYY-MM-DD>:<nonce> must start with 12 zero bits, which is 3 hex zeros), email (optional), name (optional, at most 80 characters). Returns: api_key (shown once; store it securely and send it as Authorization: Bearer), key metadata and request_id. Only the key's SHA-256 hash is stored. Cost: free. Needs no key and no payment.

      Après

      Create a free Screwtape.ai API key (10 calls per UTC day) with no human approval. Call it only when no Screwtape key is configured and you can compute SHA-256 with a code tool, because the proof of work must be computed exactly, not guessed. Do NOT call it when a key already exists, to get more quota (it is limited to 5 keys per IP per hour and per UTC day), or without a valid pow_nonce. A missing nonce returns an error that describes the challenge. Inputs: pow_nonce (required string matching ^[A-Za-z0-9_-]{8,64}$; SHA-256 of the UTF-8 string screwtape-signup-v1:<current UTC YYYY-MM-DD>:<nonce> must start with 12 zero bits, which is 3 hex zeros), email (optional), name (optional, at most 80 characters). Returns: api_key (shown once; store it securely and send it as Authorization: Bearer), key metadata and request_id. Only the key's SHA-256 hash is stored. Cost: free. Needs no key and no payment.

    • judgedescription
      Avant

      An independent pass/fail/retry verdict on ONE agent output. Call it before you return, store or act on a draft answer, especially when you have the evidence the answer should be based on. Do NOT call it to look up facts (it does no web search and only checks against the evidence you pass), to compare several agents' answers (use consensus), or for offline evaluation over a dataset. Inputs: task, agent_output, evidence, criteria, retain, and optional agent_id and version tags. Returns: verdict, score, reasons, engine, next_step, request_id, retention, the tags when sent, and receipt when signing is configured. On pass, use the output. On retry, revise using reasons and call again, at most 2 or 3 times. On fail, stop. Optional checks run first; on failure you get next_step.hint saying what to fix. Branch on next_step.action (pass, retry or escalate). Cost: 1 call. A free key covers 100 calls per UTC day; without a key, pay $0.005 USDC on Base or Solana per call with x402. Without a key…

      Après

      An independent pass/fail/retry verdict on ONE agent output. Call it before you return, store or act on a draft answer, especially when you have the evidence the answer should be based on. Do NOT call it to look up facts (it does no web search and only checks against the evidence you pass), to compare several agents' answers (use consensus), or for offline evaluation over a dataset. Inputs: task, agent_output, evidence, criteria, retain, and optional agent_id and version tags. Returns: verdict, score, reasons, engine, next_step, request_id, retention, the tags when sent, and receipt when signing is configured. On pass, use the output. On retry, revise using reasons and call again, at most 2 or 3 times. On fail, stop. Optional checks run first; on failure you get next_step.hint saying what to fix. Branch on next_step.action (pass, retry or escalate). Cost: 1 call. A free key covers 10 calls per UTC day; without a key, pay $0.005 USDC on Base or Solana per call with x402. Without a key o…

11. 24 oct. 2026, 19:45 UTC, Vérification automatique

    Nouvelle version 1.2.0 (avant : 1.1.0). Annoncée par le serveur.

12. 24 oct. 2026, 19:45 UTC, Vérification automatique

    3 outils modifiés.

    Détails, 24 oct. 2026, 19:45 UTC

    Modifiés 3

    check_duplicate_invoice consensus judge

    Les 10 entrées les plus récentes montrent tous les détails ; les plus anciennes ne montrent que les noms.

13. 24 oct. 2026, 07:28 UTC, Vérification automatique

    Début de l’observation : 6 outils.

FAQ