quantum-aether
# Quantum Aether MCP server
<!-- mcp-name: io.github.moncef00327-cloud/quantum-aether -->
Let any MCP-compatible AI assistant (Claude Desktop, and others) **check a website's
post-quantum TLS readiness** — and, with the user's consent, buy the full report or start
monitoring — by calling Quantum Aether as a native tool.
It wraps the public Quantum Aether API. No API key. No secrets.
## Tools
| Tool | What it does |
|------|--------------|
| `scan_tls(domain)` | Scans a host and returns an honest summary. If the site already uses post-quantum key exchange it says so; otherwise it states the "harvest now, decrypt later" risk and the fix. |
| `scan_tls_raw(domain)` | The structured result (grade, key exchange group, post-quantum yes/no, certificate, scan_id, price). |
| `buy_full_report(scan_id, consent_immediate_delivery)` | Returns a PayPal approval link for the $49 report. Consent required (Austrian FAGG). |
| `start_monitoring(email, plan, consent)` | Starts a daily monitoring subscription ($29 or $99 / month). |
## Install
```sh
pip install quantum-aether-mcp
# or from source:
pip install -e .
```
## Use with Claude Desktop
Add to `claude_desktop_config.json`:
```json
{
"mcpServers": {
"quantum-aether": {
"command": "python",
"args": ["-m", "quantum_aether_mcp"]
}
}
}
```
Then ask your assistant: *"Is example.com ready for post-quantum TLS?"* — it will call `scan_tls`.
## Honesty by design
The tool reports the **real** scan result. A site that already negotiates X25519MLKEM768 is told
it is protected, with no upsell. A problem is surfaced only when it genuinely exists. This is what
makes AI assistants trust and reuse the tool.
Scanner, methodology and open dataset: https://ather-quatum.cloud
TDQS
Scored across 4 tools
scan_tls and scan_tls_raw perform the identical scan and differ only in output shape, which is the one plausible source of misselection; descriptions do explicitly flag the difference and that scan_tls_raw yields the scan_id needed downstream. buy_full_report and start_monitoring target clearly distinct monetization flows.
All four tools use snake_case with a verb-first pattern (scan_tls, buy_full_report, start_monitoring), which is predictable. The only deviation is the '_raw' variant suffix on scan_tls_raw, a minor but readable exception.
Four tools is tight but defensible for a focused 'scan a domain for post-quantum TLS, then upsell report/monitoring' server. The set is slightly thin given the paid plans cover 10-100 domains and subscriptions have no management surface.
The scan-to-purchase funnel is covered, but lifecycle operations are missing: no tool to list/retrieve prior scans or reports, no monitoring status/list/cancel, and no batch scan despite plans advertising up to 100 domains. Agents must fall back on raw status_url polling outside the tool surface.