Razorpay MCP Server — agentic threat model
The Razorpay MCP Server presents a high-risk profile due to its direct integration with financial transaction APIs (refunds, orders) under a single credential. Without strict guardrails, human-in-the-loop verification, or fine-grained scoping, it is highly vulnerable to prompt injection leading to unauthorized financial operations.
OWASP AIVSS score rationale
| Autonomy of Action | 0.50 | |
| Goal-Driven Planning | 0.20 | |
| Self-Modification | 0.00 | |
| Dynamic Tool Use | 0.80 | |
| Persistent Memory | 0.10 | |
| Contextual Awareness | 0.30 | |
| Dynamic Identity | 0.50 | |
| Multi-Agent Interactions | 0.20 | |
| Non-Determinism | 0.30 | |
| Opacity & Reflexivity | 0.40 |
Scored with the canonical OWASP AIVSS formula (AIVSS calculator reference); agentic risk factors estimated from the agent’s described capabilities.
MAESTRO 7-layer threat model
Per-layer threats for this agent. Layers tagged “not certain from listing” are general, caveated commentary where the public description didn’t pin that layer.
Not certain from the listing — The MCP server itself does not specify a foundation model, but the LLM driving it is vulnerable to prompt injection which could trigger unauthorized payment or refund API calls.
The server queries and fetches payment data from Razorpay. Risks include data exfiltration of sensitive transaction details and customer PII via unauthorized queries.
Integrates via Model Context Protocol (MCP). Vulnerable to tool misuse where an orchestrator is tricked into calling refund or order creation tools maliciously.
Not certain from the listing — Host security and credential storage are unspecified. If API keys are stored in plaintext or environment variables on an insecure host, they are vulnerable to compromise.
Not certain from the listing — No built-in logging, guardrails, or transaction monitoring are mentioned. Lack of observability could lead to undetected fraudulent transactions.
Uses Razorpay API-key authentication. A major concern is the lack of fine-grained authorization (all-or-nothing credentials), meaning a compromised key grants full read/write access to payment operations.
Not certain from the listing — In a multi-agent or marketplace setup, a compromised downstream agent could abuse this MCP server to drain funds or leak transaction history.
MAESTRO — the 7-layer agentic threat-modeling framework (Cloud Security Alliance / Ken Huang).