# Serasa ClearSale Fraud Risk Analysis AI Agent Connect > Serasa ClearSale Fraud Risk Analysis MCP lets your AI client handle fraud scoring for e-commerce orders and orchestrated identity checks. Use it to submit buyer data for risk scores, manage MFA challenges, and feed chargeback data back to ClearSale to sharpen future analysis. It works across both Link de Pagamentos and Connect products to provide instant verdicts. ## Overview - **Category:** finance - **Price:** Free - **Endpoint:** https://edge.vinkius.com/vk_preview_3oE2MSybPipEtHTFm84NwxV0Hu5obrjicDwwMnTS/ai-agent-connect - **Tags:** clearsale, fraud, risk-score, chargeback, brazil, payment-link, connect, mfa ## Description You can now give your AI agent direct access to ClearSale's fraud prevention engine. This MCP handles two distinct workflows: Link de Pagamentos for standard e-commerce orders and Connect for more complex, orchestrated identity verification. When you submit an order, your agent receives a risk score from 0 to 100 and a clear verdict, such as approved, declined, or manual review. If you use the Connect workflow, your agent can even manage MFA challenges and check biolink device-binding scores. Instead of parsing messy raw data, your agent sees a clean decision line and can act immediately. You can also close the loop by reporting chargebacks or final transaction outcomes, which helps ClearSale calibrate its scoring models for your specific business needs. It handles the heavy lifting of JWT management and credential refreshing in the background, so you can focus on the decision logic. ## Tools ### get_payment_link_order This tool reads the current status, risk score, and result of a ClearSale payment-link analysis. It returns an analysisId and a 0 to 100 risk score to help your agent decide whether to release or cancel an order. ### mark_chargeback This tool registers a chargeback in the ClearSale system to improve future risk scoring. You provide the order code and notification date to help sharpen the model's detection of fraudulent buyers and devices. ### submit_payment_link_order This tool submits buyer, payment, and item data to the Link de Pagamentos API for immediate fraud scoring. It returns a status code and risk score so your agent can determine the next steps. ### update_payment_link_status This tool sends the final decision for a payment-link order back to ClearSale. Use it to report if a payment was captured or rejected so the system learns from the outcome. ### create_connect_analysis This tool triggers an orchestrated fraud analysis through a configured ClearSale Connect integration. It provides an executionId and initial decision, including potential MFA challenges or biolink scores. ### get_connect_analysis This tool retrieves the current decision, insights, and MFA state for a Connect analysis. It provides coded reasoning and device-binding scores to help your agent understand the risk verdict. ### update_connect_status This tool updates the final status of a ClearSale Connect analysis. It ensures the decision is recorded so the Connect engine can use the result as feedback for future transactions. ## Prompt Examples **Prompt:** ``` Score this order: buyer João Silva, CPF 12345678901, email joao@test.com, total BRL 500, paid by PIX, shipped to São Paulo/SP. Widget, BRL 500, qty 1. ``` **Response:** ``` The agent will call submit_payment_link_order with the buyer details and PIX payment info, then poll get_payment_link_order until the status is finalized. ``` **Prompt:** ``` Run the Connect integration INT-7 for document 12345678901 and tell me what the insights say. ``` **Response:** ``` The agent calls create_connect_analysis using the INT-7 integration ID and then uses get_connect_analysis to retrieve the decision score and coded insights. ``` **Prompt:** ``` Order ORD-42 from last month was charged back. Register it as fraud for BRL 320 on a Visa card. ``` **Response:** ``` The agent calls mark_chargeback with the order code, the dispute value, and the card brand to calibrate the risk model. ``` ## Capabilities ### Fraud Scoring Your agent receives a 0 to 100 risk score for every transaction submitted. ### MFA Management The agent can monitor and respond to MFA challenge states during Connect analyses. ### Decision Feedback Your agent reports final order outcomes to ClearSale to improve future detection accuracy. ### Identity Verification The agent accesses biolink device-binding scores to verify user identity. ### Automated Polling Your agent can track the status of pending analyses until a final verdict is reached. ## Use Cases ### Automated Order Approval An agent checks the risk score of a new PIX payment and automatically approves it if the score is low. ### High-Risk Identity Verification For sensitive transactions, the agent triggers a Connect analysis to check for MFA completion and device binding. ### Chargeback Calibration When a dispute is filed, the agent notifies ClearSale to ensure the fraudulent user is flagged in future attempts. ### Manual Review Triage The agent identifies orders marked for manual review and presents the specific coded insights to a human operator. ## Benefits - Reduces manual review time by providing clear, actionable verdicts. - Improves scoring accuracy by feeding chargeback data back into the system. - Handles complex MFA and device-binding checks through a single interface. - Simplifies API interactions by rendering raw payloads into readable decision lines. ## How It Works The process moves from initial data submission to real-time decisioning and feedback. 1. Connect your AI client to the Vinkius-hosted MCP. 2. Submit transaction or identity data using the appropriate tool. 3. Poll for the final status and risk score. 4. Review the verdict, insights, and MFA state. 5. Report the final outcome or chargeback to calibrate the model. ## Frequently Asked Questions **What is the difference between Link de Pagamentos and Connect?** Link de Pagamentos is used for scoring standard e-commerce orders. Connect is for orchestrated flows that include MFA challenges and device-binding checks. **How does the risk score work?** ClearSale provides a score from 0 to 100. A higher score indicates a higher risk of fraud. **Can I use this to reduce chargebacks?** Yes. By using the mark_chargeback tool, you provide negative feedback that helps ClearSale sharpen its scoring for specific buyers and devices. **Does the agent need to parse complex JSON responses?** No. The MCP renders responses as a clear verdict line, including the score, insights, and challenge state. **What happens if an order requires MFA?** The Connect analysis will return a pending MFA status, which your agent can monitor until the user completes the challenge. **Which product should I use for an e-commerce order?** submit_payment_link_order — it takes the buyer, the payment and the items and returns a status code plus a 0–100 score. Use Connect (create_connect_analysis) only when you have a configured integration and want its orchestrated flow with insights, MFA and biolink. **What is a good score?** ClearSale scores 0–100 where higher means riskier. The status code is the actionable signal: APA is approved, PMA routes to manual review, and rejection codes start with RP. The rendered verdict line maps this for you instead of leaving you to read the raw code. **Why should I call the status update and chargeback tools?** They are feedback, not bookkeeping. update_payment_link_status tells ClearSale which decision you actually applied and mark_chargeback reports a later chargeback; both feed the model so future scores for the same buyer, card and device signals get sharper. Skipping them silently degrades the scores you rely on.