LayerCall MCP server
Commit Nov 11, 2026Enables AI agents to perform fraud checks on IPs, emails, phone numbers, domains, and devices to assess risk scores and verdicts.
Also known as LayerCall/layercall-sdk, layercall-sdk
- Stars
- 0
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- 1.0.0
- Last commit
- Nov 11, 2026
- License
- MIT
What is LayerCall MCP?
LayerCall enables AI agents to perform fraud checks on IPs, emails, phone numbers, domains, and devices to assess risk scores and verdicts.
Summary generated from the project's README
About LayerCall MCP server
LayerCall provides a remote MCP server that allows AI agents to perform real-time fraud detection and risk assessment. Developers can connect clients like Claude Desktop, Cursor, or VS Code to the service via a URL and an API key. This setup enables an AI to evaluate the legitimacy of various identifiers during a conversation without requiring local installations.
The server exposes seven specific tools: score_ip, verify_email, lookup_phone, score_domain, score_device, verify_agent, and score_user. These tools return a risk score from 0 to 100, a verdict of allow, review, or block, and specific signals such as whether an IP is a VPN or a proxy. The service is designed to help identify bots, disposable emails, and fraudulent signups by analyzing data points like domain age, phone line types, and device fingerprints.
What you can do with it
- 1Verify email legitimacyThe agent can check if an email address is disposable or belongs to a role account.
- 2Authenticate AI agentsThe system can be used to prove that an AI agent is actually who it claims to be.
Written from the project's README. Every use case is checked against a quote from it.
How to install LayerCall MCP
Formats verified 10 Nov 2026Add to CursorAdd to GooseInstall LayerCall MCP in Claude Code
$ claude mcp add --transport http layercall-sdk https://www.layercall.com/api/mcpInstall LayerCall MCP in Cursor
~/.cursor/mcp.json
{
"mcpServers": {
"layercall-sdk": {
"url": "https://www.layercall.com/api/mcp"
}
}
}Install LayerCall MCP in VS Code
.vscode/mcp.json
{
"servers": {
"layercall-sdk": {
"type": "http",
"url": "https://www.layercall.com/api/mcp"
}
}
}Install LayerCall MCP in VS Code CLI
$ code --add-mcp '{"name":"layercall-sdk","type":"http","url":"https://www.layercall.com/api/mcp"}'Install LayerCall MCP in Codex
$ codex mcp add layercall-sdk --url https://www.layercall.com/api/mcpInstall LayerCall MCP in Codex config.toml
~/.codex/config.toml
[mcp_servers.layercall-sdk] url = "https://www.layercall.com/api/mcp"
Install LayerCall MCP in Gemini CLI
$ gemini mcp add -s user --transport http layercall-sdk https://www.layercall.com/api/mcpInstall LayerCall MCP in Gemini settings.json
~/.gemini/settings.json
{
"mcpServers": {
"layercall-sdk": {
"url": "https://www.layercall.com/api/mcp",
"type": "http"
}
}
}Configuration
LAYERCALL_API_KEY
| Name | Description | |
|---|---|---|
| LAYERCALL_API_KEY | — |
Automated checks
Ran Nov 14, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachablelayercall/layercall-sdk is public and reachablegithub.com/layercall/layercall-sdk ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)MITspdx.org/licenses/MIT.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-11-11github.com/layercall/layercall-sdk ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeinitialize answered with protocol 2025-06-18 in 918 msPass
README
View on GitHubLayerCall SDKs
Official clients for LayerCall — score an IP, email, phone number, domain, or a whole signup for fraud in a single call.
| Install | Docs | |
|---|---|---|
| Node / TypeScript | npm i layercall |
node/README.md |
| Python | pip install layercall |
python/README.md |
| CLI | npx layercall ip 8.8.8.8 |
below |
Both clients have zero dependencies. A trust check sits on your signup path, which is the worst place in an application to introduce a dependency tree.
Get a free API key — 1,000 lookups a month, no card required, no daily cap.
Try it without installing anything
export LAYERCALL_API_KEY=tl_live_...
npx layercall ip 8.8.8.8
npx layercall email someone@mailinator.com
npx layercall domain example.com
npx layercall user --ip 1.2.3.4 --email a@b.com
What it returns
Every endpoint returns the same shape: a 0–100 risk_score, an
allow | review | block verdict, and the signals behind it.
{
"ip": "185.220.101.1",
"risk_score": 60,
"verdict": "review",
"signals": {
"is_vpn": true,
"is_proxy": true,
"is_datacenter": true,
"is_tor": true,
"recent_abuse": false
},
"geo": { "country": "DE", "city": "Berlin", "asn": "AS60729" },
"vpn_provider": null
}
Two things worth reading before you integrate
Prefer step-up over rejection. The review band starts where an ordinary
commercial VPN lands, and most VPN users are ordinary customers. Send them an
OTP or a 3-D Secure challenge instead of a refusal: a real user clears it in
seconds, an attacker cannot, and a false positive costs friction rather than a
customer. This is what Stripe Radar and Sift both converged on.
null means unknown, not "no". newly_registered is null when a TLD
publishes no RDAP (.de, .ru, .ac.uk among them), and mailbox_exists is
null when the mail provider does not answer honestly — Gmail and Yahoo accept
mail for addresses that do not exist. Treating either as a negative finding is
the specific mistake these fields exist to prevent.
Endpoints
POST/GET /v1/score/ip |
VPN, proxy, Tor, datacenter, geo, ASN |
/v1/verify/email |
Syntax, MX, disposable, role account, domain age |
/v1/lookup/phone |
Numbering-plan validation worldwide, line type |
/v1/score/domain |
RDAP age, registrar, MX/SPF/DMARC, risky TLD |
/v1/score/device |
Device fingerprint reputation + bot probability |
/v1/score/user |
All of the above weighted into one verdict |
/v1/verify/agent |
Web Bot Auth (RFC 9421) — prove an AI agent is who it claims |
/v1/batch |
Up to 500 values of one type |
Full reference: layercall.com/docs · OpenAPI 3.1: layercall.com/openapi.json
Testing without spending anything
Test-mode keys return deterministic synthetic data, drawn from ranges
reserved for exactly this purpose — RFC 5737 addresses, example.com, the
555-01XX fiction block. Same shape and fields as production, so your assertions
are real ones. They never bill, never hit live data sources, and never write to
the shared reputation network.
Every fixture is documented, so you can assert on exact values rather than "did it return a number": layercall.com/docs/test-mode
MCP — call it from an AI agent
LayerCall speaks the Model Context Protocol over Streamable HTTP, so Claude Code, Claude Desktop, ChatGPT, Cursor, VS Code, Windsurf and Zed can run a fraud check mid-conversation. There is nothing to install — it is a remote server, so you add a URL and your API key:
{
"mcpServers": {
"layercall": {
"url": "https://www.layercall.com/api/mcp",
"headers": { "Authorization": "Bearer YOUR_API_KEY" }
}
}
}
Seven tools are exposed: score_ip, verify_email, lookup_phone,
score_domain, score_device, verify_agent and score_user.
VS Code names the top-level key
servers, notmcpServers. That one difference is the usual reason a copied config silently does nothing.
Per-client setup: layercall.com/docs/mcp
Framework middleware
Drop-in for the two places this usually goes. The middleware attaches
req.trust and leaves the decision to you:
import { layercall } from "layercall/express";
app.post("/signup", layercall(), (req, res) => {
if (req.trust.verdict === "block") return res.status(403).json({ error: "..." });
if (req.trust.verdict === "review") flagForManualReview(req.trust);
createAccount(req.body);
});
There is deliberately no autoBlock: true. A one-line install that starts
rejecting people is the wrong default for a fraud tool — the failure is silent,
it lands on real customers, and you find out from a support ticket.
By default it only scores POST/PUT/PATCH, because a global app.use()
bills a lookup for every request including favicon.ico.
See node/src/express.ts and
node/src/next.ts (scoreRequest, withTrust).
Guides
Written for someone mid-incident rather than someone shopping. Each one ends with what a naive version gets wrong — including the parts that need no LayerCall at all.
- How to stop fake signups
- How to block disposable email addresses at signup
- How to detect VPN and proxy users at signup
- How to stop free trial abuse
- How to add fraud checks without losing real customers
License
MIT
LayerCall MCP questions
How do I add LayerCall MCP to Claude Code?
Run claude mcp add --transport http layercall-sdk https://www.layercall.com/api/mcp in your terminal. Add --scope user to make it available in every project.
How do I add LayerCall MCP to Cursor?
Open ~/.cursor/mcp.json and add the server under mcpServers, using the Cursor config shown in the install section above.
How much does it cost to get started?
You can get a free API key that includes 1,000 lookups per month with no credit card required and no daily cap.
What information is returned in a fraud check?
Every endpoint returns a risk score from 0–100, an allow, review, or block verdict, and the specific signals behind the score.
Can I test the API without using live data or incurring costs?
Yes, test-mode keys return deterministic synthetic data that follows the same shape and fields as production without billing or hitting live sources.