Kaidn — Claude Code MCP server: fraud verdicts explained
Wires the Kaidn fraud-scoring API into your coding agent, so you can ask why a signup got blocked and what else a device has touched.
claude mcp add --env KAIDN_API_KEY=<KAIDN_API_KEY> --env KAIDN_API_URL=<KAIDN_API_URL> --env KAIDN_MCP_ALLOW_WRITES=<KAIDN_MCP_ALLOW_WRITES> --env KAIDN_MCP_MAX_QUOTA_CALLS=<KAIDN_MCP_MAX_QUOTA_CALLS> --transport stdio kaidn -- npx -y @kaidn/mcpWhat Kaidn does
Kaidn is an MCP server in front of the Kaidn fraud-scoring API. The agent gets investigation tools: rollups of verdicts and scores over a window, a list of scored events you can filter by verdict or type and search by fingerprint or user id, and a breakdown of a single event listing every check that fired with its raw figures. That is why an answer can cite evidence instead of guessing at a score.
Reads only, unless you opt in. Mutating tools appear only with --allow-writes. Config changes and GDPR subject erasure never get exposed in any mode. The API key stays in KAIDN_API_KEY and nowhere else: passed as a flag it would end up in process listings and shell history.
You need Node.js 18 or newer plus a key from the Kaidn dashboard. stdio is the default transport; remote agents go over Streamable HTTP at POST /mcp. The HTTP mode has one catch. The server listens on 127.0.0.1 and refuses to start on a broader interface until KAIDN_MCP_HTTP_TOKEN is set. Each process also gets a quota ceiling, 100 calls by default, so a looping agent cannot drain your monthly budget.
What it helps with
- work out why a signup or payment was declined by the fraud engine
- find every account sharing one device fingerprint
- clear the manual review queue first thing in the morning
- pull scored events for a user id or fingerprint
- cap quota spend during automated checks
Example prompts for the agent
- Why was this signup blocked? Break down the event in Kaidn and list the checks that fired.
- Which other accounts has this device touched, and what do they have in common?
- What is waiting in the review queue today? Pull the summary with
get_stats.
How to install Kaidn
Platform CLI
claude mcp add --env KAIDN_API_KEY=<KAIDN_API_KEY> --env KAIDN_API_URL=<KAIDN_API_URL> --env KAIDN_MCP_ALLOW_WRITES=<KAIDN_MCP_ALLOW_WRITES> --env KAIDN_MCP_MAX_QUOTA_CALLS=<KAIDN_MCP_MAX_QUOTA_CALLS> --transport stdio kaidn -- npx -y @kaidn/mcpReplace <…> with your own values
Platform CLI
codex mcp add kaidn --env KAIDN_API_KEY=<KAIDN_API_KEY> --env KAIDN_API_URL=<KAIDN_API_URL> --env KAIDN_MCP_ALLOW_WRITES=<KAIDN_MCP_ALLOW_WRITES> --env KAIDN_MCP_MAX_QUOTA_CALLS=<KAIDN_MCP_MAX_QUOTA_CALLS> -- npx -y @kaidn/mcpReplace <…> with your own values
Config file
{
"mcpServers": {
"kaidn": {
"command": "npx",
"args": [
"-y",
"@kaidn/mcp"
],
"env": {
"KAIDN_API_KEY": "<KAIDN_API_KEY>",
"KAIDN_API_URL": "<KAIDN_API_URL>",
"KAIDN_MCP_ALLOW_WRITES": "<KAIDN_MCP_ALLOW_WRITES>",
"KAIDN_MCP_MAX_QUOTA_CALLS": "<KAIDN_MCP_MAX_QUOTA_CALLS>"
}
}
}
}Replace <…> with your own values
Config file
{
"servers": {
"kaidn": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@kaidn/mcp"
],
"env": {
"KAIDN_API_KEY": "<KAIDN_API_KEY>",
"KAIDN_API_URL": "<KAIDN_API_URL>",
"KAIDN_MCP_ALLOW_WRITES": "<KAIDN_MCP_ALLOW_WRITES>",
"KAIDN_MCP_MAX_QUOTA_CALLS": "<KAIDN_MCP_MAX_QUOTA_CALLS>"
}
}
}
}Replace <…> with your own values
Platform CLI
gemini mcp add -e KAIDN_API_KEY=<KAIDN_API_KEY> -e KAIDN_API_URL=<KAIDN_API_URL> -e KAIDN_MCP_ALLOW_WRITES=<KAIDN_MCP_ALLOW_WRITES> -e KAIDN_MCP_MAX_QUOTA_CALLS=<KAIDN_MCP_MAX_QUOTA_CALLS> kaidn npx -y @kaidn/mcpReplace <…> with your own values
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about Kaidn
How do I install Kaidn in Claude Code?
Run claude mcp add kaidn-mcp -e KAIDN_API_KEY=<KAIDN_API_KEY> -- npx -y @kaidn/mcp. The longer form adds the API URL, write access and quota limit through --env flags plus --transport stdio.
What does Kaidn do?
It exposes Kaidn's fraud scoring to the agent: verdict rollups, event lists, and a per-event explanation carrying the numbers behind each check. The model can then justify a verdict with data.
What do I need before running it?
Node.js 18 or later and an API key from the Kaidn dashboard, supplied via KAIDN_API_KEY. A quota call limit per process is on by default.
