Skip to content
MCP server

scamgate

Offline allow, ask or deny check for links and payments an AI agent is about to act on

Runs locallyStatus and history
Install
Pricing
Open source
Platforms
CLI, self-hosted
Runs
Locally
Version
0.3.3
Listing updated
2 days ago

About scamgate

scamgate is an open-source MCP server for security. It runs locally on your machine and is started by your AI client.

Scamgate scores a link or a payment with deterministic rules and returns allow (under 25 points), ask (25 to 69) or deny (70 or more), with a plain-English reason for each rule that fired. There is no language model in the loop and no network call during a check, so the same input always gets the same verdict. Link rules cover known-bad feeds, brand look-alikes and homoglyphs, addresses hidden before an @, raw IP hosts, punycode and pages asking for a seed phrase. Payment rules look at new payees plus urgency, changed bank details and irreversible methods such as crypto or gift cards.

Its distinctive claim is enforcement. Over MCP the agent chooses whether to ask, and an injected prompt could tell it not to, so a hook for Claude Code and Hermes runs the check in the harness before every gated tool call and blocks on deny. The hook fails closed if it crashes and blocks edits to its own files. Payee history is stored locally as salted hashes.

The README says it is advisory and can miss scams. Python 3.11 or newer is required, feeds are downloaded on request, and automatic refresh is off by default.

Install

Run in a terminal

claude mcp add scamgate --env SCAMGATE_URLHAUS_AUTH_KEY='<SCAMGATE_URLHAUS_AUTH_KEY>' -- uvx scamgate

Add to claude_desktop_config.json

{
  "mcpServers": {
    "scamgate": {
      "command": "uvx",
      "args": [
        "scamgate"
      ],
      "env": {
        "SCAMGATE_URLHAUS_AUTH_KEY": "<SCAMGATE_URLHAUS_AUTH_KEY>"
      }
    }
  }
}

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "scamgate": {
      "command": "uvx",
      "args": [
        "scamgate"
      ],
      "env": {
        "SCAMGATE_URLHAUS_AUTH_KEY": "<SCAMGATE_URLHAUS_AUTH_KEY>"
      }
    }
  }
}

Add to .vscode/mcp.json

{
  "servers": {
    "scamgate": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "scamgate"
      ],
      "env": {
        "SCAMGATE_URLHAUS_AUTH_KEY": "<SCAMGATE_URLHAUS_AUTH_KEY>"
      }
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "scamgate": {
      "command": "uvx",
      "args": [
        "scamgate"
      ],
      "env": {
        "SCAMGATE_URLHAUS_AUTH_KEY": "<SCAMGATE_URLHAUS_AUTH_KEY>"
      }
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "scamgate": {
      "command": "uvx",
      "args": [
        "scamgate"
      ],
      "env": {
        "SCAMGATE_URLHAUS_AUTH_KEY": "<SCAMGATE_URLHAUS_AUTH_KEY>"
      }
    }
  }
}

Configuration

  • SCAMGATE_AUTO_UPDATESet to 1 to refresh the phishing lists every 6 hours while the server runs (off by default).
  • SCAMGATE_URLHAUS_AUTH_KEYOptional abuse.ch Auth-Key that adds the URLhaus malware-link list.

Tools this server provides (2)

Taken from the project documentation; the running server may differ.

check_url
Scores a link and returns allow, ask or deny with reasons.
check_payment
Scores a payment and returns allow, ask or deny with reasons.

What it does

Deterministic scoring
Rules add points and the total maps to allow, ask or deny.
Link checks
Detects look-alike brands, hidden hosts, raw IPs and credential prompts.
Payment checks
Flags new payees, changed bank details and irreversible methods.
Enforcing hook
Runs in the harness so the agent cannot skip the check.
Fails closed
A crash or malformed input blocks the call.
Local payee history
Remembers who you have paid, stored as salted hashes.

Good for

  • Stop an agent opening a typosquatted login page
  • Gate payments to a payee that changed bank details
  • Add a safety hook to Claude Code or Hermes
  • Review why a link scored deny

Pricing

Open source

MIT licence, self-hosted

Prices change. Check odaiameera.github.io before you buy.

scamgate FAQ

Does it call the network when checking?

No. The site says there are no network calls while checking, and no LLM judging.

What are the score thresholds?

Under 25 allow, 25 to 69 ask, 70 or more deny.

Is it reliable?

The README calls it advisory: it can miss scams and flag legitimate actions.

How is it licensed?

Open source under the MIT licence and self-hosted.

What does it need?

Python 3.11 or newer; install from PyPI or a clone.

Is scamgate free?

Yes. scamgate is open source, so you can use and self-host it for free.

How do I add scamgate to Claude Code?

Run this command in your terminal: claude mcp add scamgate --env SCAMGATE_URLHAUS_AUTH_KEY='<SCAMGATE_URLHAUS_AUTH_KEY>' -- uvx scamgate

What are the best alternatives to scamgate?

The closest alternatives to scamgate are REA, Agent Roast, Midplane, Leash and Tinplate Edge AI Safety, Security & EU Compliance.

Alternatives to scamgate

  1. REA

    MCP server

    CLI and MCP server that lets coding agents reverse engineer apps and binaries

    Runs locally
    • Open source
    • 464K GitHub stars
    • Developer tools
  2. Agent Roast

    MCP server

    Offline review that flags agent changes which weaken tests, assertions or CI

    Runs locally
    • Open source
    • 4 Hacker News points
    • Developer tools
  3. Midplane

    MCP server

    Postgres gateway for AI agents with table grants, column masking and approval for writes

    Runs locally
    • Open source
    • 5 GitHub stars
    • Databases
  4. Leash

    MCP server

    Review what your AI touched, approve risky actions and undo supported changes

    • 1 tool
    • Remote
    • Free
    • 1 GitHub star
    • Security
  5. SecureCloud

    MCP server

    SentinelOne and IT security catalog, prices, quotes and checkout links for Israel

    • 7 tools
    • Remote
    • Free
    • Security