Skip to main content
Glama
pathguard-mcp

pathguard-mcp

by Utee

pathguard-mcp

PathGuard MCP lets MCP-compatible AI agents check crypto transactions for scams and mistakes directly in conversation. It supports local MCP clients such as Claude Desktop, Claude Code, Gemini CLI, Grok CLI, Cursor, and Codex, plus remote Streamable HTTP deployments.

MCP Registry

Listed on the official MCP Registry as org.cieltech.pathguard/pathguard.

The production MCP server is a thin, authenticated interface to PathGuard's transaction-safety API. PathGuard analyzes transactions before a send; it does not sign, broadcast, or transfer funds.

Production MCP

Remote Streamable HTTP endpoint:

https://api.pathguard.cieltech.org/mcp/

Remote clients use OAuth to connect an existing PathGuard account. During authorization, PathGuard verifies the user's existing API key. The API key is not exposed to the AI client or returned through MCP.

For local/stdio use, the API key is supplied to the local server process through PATHGUARD_API_KEY.

Related MCP server: rugradar

Install

The package is currently installed directly from this public repository and is not published to PyPI yet.

git clone https://github.com/Utee/pathguard-mcp.git
cd pathguard-mcp
python -m pip install .

For development, use an editable install:

python -m pip install -e .

Requires Python 3.10+ and mcp >= 2.0.

Get an API key

Create a PathGuard account and API key from the PathGuard API documentation.

Keep your API key private. Do not commit it to GitHub or expose it in browser-side code.

Local MCP clients

For local clients, the server uses stdio and authenticates to the PathGuard API with PATHGUARD_API_KEY.

Claude Desktop

Add this to your Claude Desktop config:

{
  "mcpServers": {
    "pathguard": {
      "command": "pathguard-mcp",
      "env": {
        "PATHGUARD_API_KEY": "pg_your_key_here"
      }
    }
  }
}

Claude Code

claude mcp add --transport stdio pathguard -- pathguard-mcp

Set PATHGUARD_API_KEY in the environment that starts Claude Code.

Gemini CLI

{
  "mcpServers": {
    "pathguard": {
      "command": "pathguard-mcp",
      "env": {
        "PATHGUARD_API_KEY": "pg_your_key_here"
      }
    }
  }
}

Grok CLI

export PATHGUARD_API_KEY=pg_your_key_here
grok mcp add pathguard -- pathguard-mcp

Codex, Cursor, VS Code, and other local MCP clients

Use the same stdio server definition as the Claude Desktop example. The exact configuration location varies by client.

Remote clients

For remote clients, use the production OAuth-protected Streamable HTTP endpoint shown above. Never put a PathGuard API key in browser-side configuration.

ChatGPT

In a ChatGPT workspace with Developer Mode enabled, create a custom MCP app and provide:

https://api.pathguard.cieltech.org/mcp/

Choose OAuth authentication. ChatGPT will use the PathGuard authorization flow to connect an existing PathGuard account.

The authorization flow verifies the user's PathGuard API key, then issues OAuth credentials for the MCP connection. The raw PathGuard API key is not returned to ChatGPT.

Available tools

Tool

What it does

check_transaction

Inspect one crypto destination and optional amount for PathGuard risk signals.

preflight_transaction

Run the full pre-send check and return ALLOW, REVIEW, or BLOCK.

check_transactions_batch

Check up to 1,000 transactions and return a compact risk summary or full results.

report_scam_address

Record a suspected scam address when the user explicitly asks to report it.

get_usage_status

View the connected account's plan and monthly scan usage.

get_profile

Identify the PathGuard account represented by the OAuth connection.

Tool safety boundary

PathGuard MCP is analysis-only. Its tools do not:

  • sign transactions;

  • broadcast transactions;

  • transfer funds;

  • request private keys or seed phrases;

  • replace your application's authorization controls.

A BLOCK result is a PathGuard safety signal. The wallet, application, or agent remains responsible for authorization, signing, and final transaction execution.

Batch checking

check_transactions_batch supports up to 1,000 items.

By default it returns a compact summary containing:

  • total items;

  • processed items;

  • ALLOW / REVIEW / BLOCK counts;

  • maximum risk score;

  • flag counts;

  • a sample of flagged transactions.

Use detail_level="full" when individual results are required.

Security

  • Keep API keys server-side and private.

  • Use HTTPS for remote deployments.

  • OAuth protects the production remote MCP connection.

  • The production MCP server maps the authenticated OAuth subject to an existing PathGuard account.

  • Internal server-to-server authentication is not exposed as an MCP argument or model-visible credential.

  • Never commit API keys, production credentials, or local .env files.

Full API reference

See the PathGuard API documentation for endpoint schemas, rate limits, pricing, and risk-flag information.

PathGuard MCP: https://pathguard.cieltech.org/mcp

Moving existing connections to the custom domain

Remote OAuth clients previously connected to the Railway address must create a new connection using https://api.pathguard.cieltech.org/mcp/ and authorize again with their existing PathGuard API key. The old Railway endpoint is in maintenance; imported OAuth records do not automatically update a client's saved server URL or issuer.

For local stdio clients, update your checkout (git pull and python -m pip install .) or explicitly set PATHGUARD_API_BASE=https://api.pathguard.cieltech.org alongside PATHGUARD_API_KEY, then restart the MCP client. Existing API keys remain valid.

Optional token and deposit details

Existing tool names are unchanged. check_transaction and each check_transactions_batch item accept an optional transfer object:

{"address":"TAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA","chain":"TRON","transfer":{"token":"USDC","token_kind":"token","network":"TRON","destination_type":"exchange","exchange":"Kraken"}}

Exact networks: ETHEREUM, BASE, ARBITRUM, OPTIMISM, BSC, POLYGON, BITCOIN, SOLANA, TRON, XRPL, STELLAR. The address family remains EVM/BTC/SOL/TRON/XRP/XLM. Optional transfer fields include token_contract, recipient_network, memo and memo_required. Use native only for an actual native coin, not a wrapped token. The response includes compatibility reasons and official-source review dates where available. Selected exchange rules cover Kraken, Coinbase consumer accounts and Binance.com; Coinbase Exchange, Coinbase Wallet, Binance.US, unverified combinations and personal wallet acceptance remain Unknown—confirm with the recipient. OFAC screening covers EVM/BTC/SOL/TRON only. Only selected USDC/USDT mainnet identifiers are compared with dated issuer references. The actual wallet transaction, memo account ownership and live deposit availability are not verified.

Update your GitHub-installed package and restart your MCP client to refresh its tool schemas. This source update does not publish a package release. The hosted OAuth MCP at https://api.pathguard.cieltech.org/mcp/ also extends its existing preflight_transaction tool with the same transfer object. Existing address-only calls still work.

Issuer-reference token identity

Pass the contract/mint from the sending wallet's asset details in transfer.token_contract. Mainnet coverage: USDC on Ethereum, Base, Arbitrum One, OP Mainnet, Polygon PoS and Solana; USDT on Ethereum, TRON and Solana. References come from Circle and Tether and expire after seven days unless reviewed again.

Read compatibility.token_identity for matched, mismatch or unknown, source dates and limitations. Matched is an identifier comparison, not proof of recipient acceptance or which asset a wallet will actually send. A mismatch may represent a legitimate wrapped or bridged asset, not necessarily a scam. Missing, stale, malformed or uncovered identifiers remain unknown. The existing tool names and arguments are unchanged.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      A
      quality
      C
      maintenance
      Scans suspicious messages, URLs, and text for scams inside any MCP-compatible AI assistant. No signup or API key needed for anonymous use.
      1
      56 npm
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      Enables AI assistants to check whether a crypto token, link, or forwarded "gem" message is a scam by running contract rules, live test trades, creator history, and market data across 64 networks, returning plain-English or Pidgin verdicts. Also supports scanning messages, reviewing a token's past checks, and explaining individual findings.
      5
      MIT
    • A
      license
      Not graded
      quality
      C
      maintenance
      Enables AI agents to check a payment recipient against the sender's own on-chain history before funds move, flagging address-poisoning look-alikes, spoofed 'from you' transfers, and fake tokens via the Blockscout PRO API. It also scans a wallet for poisoning attempts aimed at it, returning a verdict of OK, CAUTION, or STOP.
      MIT
    • A
      license
      A
      quality
      B
      maintenance
      Enables users and AI agents to verify whether a crypto recipient address is an address-poisoning lookalike or has an invalid EIP-55 checksum before sending funds, and provides access to an on-chain history scan for suspicious transfers.
      2
      MIT