About AI Sovereignty Scan Mcp
AI Sovereignty Scan Mcp is an MCP server for checking vendors’ data jurisdictions and legal exposure under the EU AI Act context. It helps teams review providers individually or across a stack, with provider listings and EU-based alternatives.
Key features
- 01Vendor jurisdiction profiles
- 02Stack-wide exposure summaries
- 03Provider list with category filters
- 04US CLOUD Act provider list
- 05EU-based alternative suggestions
Use cases
- 01Review a vendor’s data residency and legal framework
- 02Count CLOUD Act exposure across a provider stack
- 03Find EU alternatives in the same category
MCP support
Exposes sovereignty scanning tools to MCP clients, including provider and stack scans, provider listings, and EU alternative suggestions. It is available as an HTTP MCP endpoint.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://sovereignty-scan.kajaril.com/mcpclaude mcp add --transport http sovereignty-scan https://sovereignty-scan.kajaril.com/mcp{"mcpServers":{"sovereignty-scan":{"type":"http","url":"https://sovereignty-scan.kajaril.com/mcp"}}}Open in Cursor{"servers":{"sovereignty-scan":{"type":"http","url":"https://sovereignty-scan.kajaril.com/mcp"}}}Open in VS Code{"mcpServers":{"sovereignty-scan":{"type":"http","url":"https://sovereignty-scan.kajaril.com/mcp"}}}{"mcpServers":{"sovereignty-scan":{"serverUrl":"https://sovereignty-scan.kajaril.com/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
Details
From the project README@kajaril/sovereignty-scan-mcp
MCP server for EU AI Act vendor sovereignty scanning. MIT-licensed free tier.
Know where your stack processes data before the enforcer does. Covers 55 providers across 12 categories.
Install
1. Get a free API key
curl -s -X POST https://sovereignty-scan.kajaril.com/register \
-H "Content-Type: application/json" \
-d '{"email":"[email protected]"}' | jq .api_key
Save the returned key — it cannot be recovered.
2. Add to claude_desktop_config.json
{
"mcpServers": {
"sovereignty-scan": {
"type": "http",
"url": "https://sovereignty-scan.kajaril.com/mcp",
"headers": {
"Authorization": "Bearer ks_free_YOUR_KEY"
}
}
}
}
Restart Claude Desktop.
Client compatibility
| Client | Status |
|---|---|
| Claude Desktop | ✓ Supported |
| Cursor / Windsurf | ✓ Supported (HTTP MCP) |
| claude.ai web | ✗ Not supported (no HTTP MCP) |
Quick test
Verify the endpoint is live:
curl -s https://sovereignty-scan.kajaril.com/health | jq .
Call a tool with your key:
curl -s -X POST https://sovereignty-scan.kajaril.com/mcp \
-H "Authorization: Bearer ks_free_YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"scan_provider","arguments":{"name":"cloudflare"}}}' \
| jq .
Tools
scan_provider — Full jurisdictional profile for a single vendor: headquarters country, data residency regions, EU residency option, US CLOUD Act exposure, GDPR DPA availability, and legal framework.
name— string, case-insensitive
Example response:
{
"name": "Cloudflare",
"hq_country": "US",
"data_residency_regions": ["US", "EU", "APAC"],
"eu_residency_option": true,
"us_cloud_act_subject": true,
"gdpr_dpa_available": true,
"legal_framework": "GDPR+SCC"
}
scan_stack — Aggregate jurisdictional summary for a list of vendors: CLOUD Act exposure count, EU residency coverage, missing DPAs. Maximum 50 providers per call.
providers— string[], max 50
list_providers — List all tracked providers. Optional category filter.
category?— AI · Hosting · Database · Auth · Analytics · Observability · CI/CD · Communications · Payments · Search · Sandbox · Cache
get_us_cloud_act_providers — All providers subject to US CLOUD Act compelled disclosure (18 U.S.C. § 2713). No parameters.
suggest_eu_alternatives — EU/EEA/UK/CH-based alternatives in the same category as a given provider. Deterministic ordering: EU/EEA first, then UK/CH. Capped at 10.
provider_name— string, case-insensitive
Pricing
| Free | Paid | |
|---|---|---|
| Price | — | €39–149 / mo |
| License | MIT | Subscription |
| Status | Live | Coming soon |
| Output | Jurisdiction, residency, legal framework, CLOUD Act | + Proprietary risk score + Remediation guidance |
| Auth | API key (free registration) | API key |
| Rate limit | 100 req / day / IP | Extended |
Paid tier notifications: [email protected]
Self-hosting
Requires a Cloudflare account (Workers + D1 + KV).
1. Clone and install
git clone https://github.com/mightbesaad/sovereignty-scan-mcp
cd sovereignty-scan-mcp
npm install
2. Create infrastructure
npx wrangler d1 create sovereignty-db-free
npx wrangler kv namespace create CACHE_KV
npx wrangler kv namespace create KEYS_KV
npx wrangler rate-limit:namespace create RATE_LIMITER
npx wrangler rate-limit:namespace create BURST_LIMITER
Copy the IDs printed by each command into wrangler.jsonc under d1_databases, kv_namespaces, and unsafe.bindings.
3. Apply schema and seed data
npx wrangler d1 execute sovereignty-db-free --remote --file=migrations/0001_providers.sql
node --input-type=module -e "
import { generateSeedSQL } from './src/seed.js';
process.stdout.write(generateSeedSQL());
" | npx wrangler d1 execute sovereignty-db-free --remote --command=-
4. Deploy
npx wrangler deploy
The custom domain (sovereignty-scan.kajaril.com) in the default config is owned by kajaril — remove or replace the routes entry with your own domain or use the default *.workers.dev URL.
Health
GET https://sovereignty-scan.kajaril.com/health
Returns a structured payload (HTTP 200):
{
"status": "ok",
"provider_count": 55,
"anthropic_path_count": 3,
"last_kv_refresh": "2026-05-11T00:00:00.000Z",
"cache_age_seconds": 86400,
"schema_version": "0001"
}
status is "ok" when D1 is reachable, "degraded" otherwise. cache_age_seconds is null if the KV cache has never been warmed.
License
MIT — see LICENSE.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.