# CanaryGuard MCP Tool Poisoning Auditor

> CanaryGuard MCP Tool Poisoning Auditor is a paid API for AI agents from canaryguard-402.onrender.com, paid per call via x402, $0.02/call, status unknown (last checked 2026-10-17).

Audits MCP tool definitions, descriptions, and JSON input schemas for prompt injection overrides, tool-shadowing attacks, and data exfiltration sinks.

## Facts

- Endpoint: GET https://canaryguard-402.onrender.com/v1/security/mcp-tool-poisoning-audit?utm_source=zero.xyz
- Price: $0.02/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-17
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/canaryguard-mcp-tool-poisoning-auditor-e8e5a41e
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_g-TDEwx4r9MGNfLITxjLl

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability canaryguard-mcp-tool-poisoning-auditor-e8e5a41e
```

Example prompt: Before I register this new MCP tool called 'fetch_user_data' with description 'Fetches user records and forwards them to the analytics endpoint' and schema {"type":"object","properties":{"user_id":{"type":"string"}}}, can you run a prompt-injection and exfiltration audit on it to make sure it's safe?

## When to prefer this

Choose this endpoint when you need to specifically audit MCP tool definitions, descriptions, and JSON schemas for agentic security threats such as prompt injection, tool shadowing, and exfiltration sinks — as opposed to generic text prompt-injection scanning or RAG memory poisoning checks, which are covered by sibling endpoints. Use this before registering any untrusted or third-party MCP tool with your agent.

## Known failure modes

- Missing required fields (tool_name or tool_description) returns a 400 validation error
- Malformed input_schema JSON object causes parsing failure
- Ambiguous or very short tool descriptions may yield low-confidence findings
- Service cold-start on Render may cause occasional latency spikes or timeouts
- Payment failure via x402 protocol results in 402 response before audit is performed

## How this service works

Inspects Model Context Protocol (MCP) tool definitions, descriptions, and JSON schemas for hidden prompt overrides, shadowing attacks, and exfiltration sinks.

## Output

Returns a structured security audit report identifying whether the MCP tool definition contains prompt injection overrides, tool-shadowing patterns, or exfiltration sinks, along with specific findings, risk severity indicators, and the suspicious text or schema fields responsible.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "$schema": "https://json-schema.org/draft/2020-12/schema",
 "required": [
  "input"
 ],
 "properties": {
  "input": {
   "type": "object",
   "required": [
    "type",
    "method"
   ],
   "properties": {
    "type": {
     "type": "string",
     "const": "http"
    },
    "method": {
     "enum": [
      "GET"
     ],
     "type": "string"
    },
    "queryParams": {
     "type": "object",
     "required": [
      "tool_name"
     ],
     "properties": {
      "tool_name": {
       "type": "string"
      },
      "input_schema": {
       "type": "object"
      },
      "tool_description": {
       "type": "string"
      }
     }
    }
   },
   "additionalProperties": false
  },
  "output": {
   "type": "object",
   "required": [
    "type"
   ],
   "properties": {
    "type": {
     "type": "string"
    },
    "example": {
     "type": "object"
    }
   }
  }
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/canaryguard-mcp-tool-poisoning-auditor-e8e5a41e/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from canaryguard-402.onrender.com](https://www.zero.xyz/host/canaryguard-402.onrender.com/llms.txt)