Security audit of AI-agent session records: 25 checks, three-level verdict, per-check evidence.
Maintain this server? Claim it
Call this server's tools from your browser. Nothing to install.
Requests go through mcptry because browsers can't call most MCP servers directly. Header values are sent to the server for that request only and are never stored or logged.
Tool list from our last check. Connect live to refresh it.
Full deterministic security audit of an AI-agent session record (25 checks, three-level verdict: verified / unverified / unknown, per-check evidence). Provide the record as {profile, responses}, or as {session: {profile, responses}}, or as {question: "<session record JSON string>"}. Never include passwords, private keys, seed phrases, or production credentials — such input is rejected before execution.
Agent profile object of the session record.
Whole session record object ({profile, responses}).
Session record as a JSON string (same form as the HTTP checkup endpoint).
Responses object of the session record, keyed by check id.
This tool isn't marked read-only, so it may change something on the server or in an account you connected.
Pick your client. Values in capitals, like YOUR_TOKEN, are yours to fill in.
Run in a terminal
claude mcp add --transport http moyan-audit https://aipay.sixu-ai.net.cn/mcp
Full deterministic security audit of an AI-agent session record (25 checks, three-level verdict: verified / unverified / unknown, per-check evidence). Provide the record as {profile, responses}, or as {session: {profile, responses}}, or as {question: "<session record JSON string>"}. Never include passwords, private keys, seed phrases, or production credentials — such input is rejected before execution.
| Argument | Type | Description |
|---|---|---|
| profile | object | Agent profile object of the session record. |
| session | object | Whole session record object ({profile, responses}). |
| question | string | Session record as a JSON string (same form as the HTTP checkup endpoint). |
| responses | object | Responses object of the session record, keyed by check id. |
{
"type": "object",
"properties": {
"profile": {
"type": "object",
"description": "Agent profile object of the session record."
},
"session": {
"type": "object",
"description": "Whole session record object ({profile, responses})."
},
"question": {
"type": "string",
"description": "Session record as a JSON string (same form as the HTTP checkup endpoint)."
},
"responses": {
"type": "object",
"description": "Responses object of the session record, keyed by check id."
}
},
"additionalProperties": true
}Single-event security pre-check from a free-text incident description (deterministic feature pre-check; patterns reported by the submitter are marked unverified, unaddressed categories unknown). If the description embeds a structured session record JSON, the full 25-check engine runs instead. Never include passwords, private keys, seed phrases, or production credentials — such input is rejected before execution.
| Argument | Type | Description |
|---|---|---|
| text* | string | Free-text incident description. |
| title | string | Optional incident title. |
| context | string | Optional additional context. |
| endpoint | string | Optional target endpoint involved. |
{
"type": "object",
"required": [
"text"
],
"properties": {
"text": {
"type": "string",
"description": "Free-text incident description."
},
"title": {
"type": "string",
"description": "Optional incident title."
},
"context": {
"type": "string",
"description": "Optional additional context."
},
"endpoint": {
"type": "string",
"description": "Optional target endpoint involved."
}
},
"additionalProperties": false
}We read every tool's name, description and inputs the way the model will, looking for instructions hidden from you. A clean result means no warning signs, not a guarantee. Scanned Nov 22, 2026.
No readOnlyHint / destructiveHint annotations, so clients can't tell a read from a write without asking.
Fix: Add annotations to each tool, e.g. { readOnlyHint: true } for tools that only read.