The full upstream README, mirrored here for reference. Install config, tool schemas, adoption signals, and an original overview live on the Pathguard listing page.
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.
Listed on the official MCP Registry as
org.cieltech.pathguard/pathguard.
The MCP server is a thin wrapper around the authenticated PathGuard API. Your PathGuard API key remains the credential used by the server and should never be exposed to an AI client or committed to source control.
The package is currently installed directly from this public repository and is not published to PyPI yet.
For development, use an editable install:
Requires Python 3.10+ and mcp >= 2.0.
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.
After installation, verify the MCP server starts with the default stdio transport:
The stdio process waits for MCP messages from the client. This is normal; desktop and CLI MCP clients manage the process for you.
To point the MCP server at a local PathGuard API instead of production:
Install this package wherever the client runs, then give the server your PathGuard API key. The examples below use the installed pathguard-mcp command. If it is not on your PATH, replace it with the full executable path.
Add this to your Claude Desktop config (claude_desktop_config.json):
Restart Claude Desktop, and Claude will have access to PathGuard's tools automatically.
Set PATHGUARD_API_KEY in the environment that starts Claude Code. Alternatively, use Claude Code's MCP configuration to pass that variable explicitly.
Add this entry to ~/.gemini/settings.json for a user-wide setup, or .gemini/settings.json in a project:
Or add it from the command line after exporting your key:
After exporting your key, add the local stdio server:
To store the configuration manually in ~/.grok/config.toml, use:
Use the same stdio server definition as the Claude Desktop example: command pathguard-mcp, no arguments, and PATHGUARD_API_KEY in its environment. The exact configuration-file location and UI differ by client; the MCP server settings are the same.
For browser-based MCP clients such as ChatGPT or Grok, deploy PathGuard MCP with Streamable HTTP behind HTTPS. Never put a PathGuard API key in client-side browser configuration.
Start the HTTP server locally or in your deployment environment:
The MCP endpoint is http://<host>:8000/mcp. In production, publish it as a protected HTTPS URL such as https://mcp.example.com/mcp.
The server also supports deployment configuration through environment variables:
PATHGUARD_MCP_ALLOWED_HOSTS accepts a comma-separated list. PORT is used when PATHGUARD_MCP_PORT is not set.
In a ChatGPT workspace with Developer Mode enabled, create a custom MCP app, provide your public https://.../mcp endpoint, scan its tools, and save or publish it as appropriate for your workspace. ChatGPT connects to remote MCP servers for this use case.
Use Grok's custom connector flow and provide your public https://.../mcp endpoint. The server must be reachable over HTTPS.
PATHGUARD_API_KEY server-side and private..env files.| Tool | What it does |
|---|---|
check_transaction | Scan a single transaction for scam addresses, clipboard-hijack patterns, and typo mistakes |
check_transactions_batch | Scan up to 1,000 transactions in one call |
report_scam_address | Flag an address as a scam (community reporting) |
get_usage_status | Check your current plan, quota, and usage this month |
See the PathGuard API documentation for complete endpoint documentation, rate limits, and flag meanings. This MCP server is a thin wrapper over the same API.