- Lucairn Privacy Gateway
Lucairn Privacy Gateway
Lucairn Privacy Gateway — MCP Server
EU-based privacy gateway for AI. Sits between your MCP-enabled agent and the upstream LLM (Claude, GPT-4o, o1/o3/o4) and automatically removes PII before the model ever sees it. Each response carries a signed compliance certificate proving what was redacted.
Free Developer tier: 1,000 requests/month. Cross-provider BYOK supported (one or both of ANTHROPIC_API_KEY and OPENAI_API_KEY).
Quick start
npx -y @lucairn/mcp-server
The server speaks MCP over stdio and exposes one tool, chat_via_lucairn, using the Anthropic Messages API wire format.
Configure your MCP client
Claude Desktop / Cline / Continue / Cursor
Add this to your MCP config file (claude_desktop_config.json, cline_mcp_settings.json, etc.):
{
"mcpServers": {
"lucairn": {
"command": "npx",
"args": ["-y", "@lucairn/mcp-server"],
"env": {
"LUCAIRN_API_KEY": "<your_lucairn_api_key>",
"ANTHROPIC_API_KEY": "<optional_byok_anthropic_key>",
"OPENAI_API_KEY": "<optional_byok_openai_key>"
}
}
}
}
Restart your client. The chat_via_lucairn tool becomes available immediately.
Get a Lucairn API key
Free Developer tier — sign up at https://lucairn.eu/account/signup. No credit card required. 1,000 requests/month included.
What the server does
- Your agent calls
chat_via_lucairnwith an Anthropic-Messages-API-shaped payload (model, messages, max_tokens, optional system). - The Lucairn gateway runs three layers of PII detection on user content:
- Layer 1: Known-entity matching (your tenant's named entities)
- Layer 2: Presidio NER (names, emails, IBANs, addresses, phone numbers, customer IDs, …)
- Layer 3: GPU-hosted custom-trained PII shield (Pro/Enterprise tier)
- Detected PII is replaced with placeholders (
[PERSON_1],[EMAIL_2],[IBAN_3], etc.) before reaching the upstream LLM. - The upstream model (Claude or OpenAI) sees only sanitized text. It never sees raw PII.
- The response comes back with a signed compliance certificate (ed25519 + RFC 3161 timestamp + Sigstore Rekor inclusion proof) that proves what was redacted, when, and by which sanitizer layer.
- Pro and Enterprise tiers can re-link placeholders to original values in the response. Developer (free) tier returns placeholders verbatim — useful for testing the redaction surface.
Provider routing
The gateway picks the upstream provider from the model parameter:
| Model prefix | Routes to | Env var used |
|---|---|---|
claude-*, anthropic-* | Anthropic | ANTHROPIC_API_KEY |
gpt-*, openai-*, o1-*, o3-*, o4-* | OpenAI | OPENAI_API_KEY |
If you only set one of the two BYOK env vars, you can only call models from that provider. If you set neither, calls use Lucairn's built-in upstream credits (counted against your quota).
Links
- Source: https://github.com/Declade/lucairn-sdks (mcp-server lives at
mcp-server/) - npm: https://www.npmjs.com/package/@lucairn/mcp-server
- Glama listing: https://glama.ai/mcp/servers/Declade/lucairn-sdks
- Docs: https://lucairn.eu
- Sign up (free): https://lucairn.eu/account/signup
License
MIT — see LICENSE.
Server Config
{
"mcpServers": {
"lucairn": {
"command": "npx",
"args": [
"-y",
"@lucairn/mcp-server"
],
"env": {
"LUCAIRN_API_KEY": "<YOUR_LUCAIRN_API_KEY>",
"ANTHROPIC_API_KEY": "<OPTIONAL_BYOK_ANTHROPIC_KEY>",
"OPENAI_API_KEY": "<OPTIONAL_BYOK_OPENAI_KEY>"
}
}
}
}