Skip to content

Security: gtorreal/buda-mcp

Security

SECURITY.md

Security Policy

Supported Versions

Only the latest published version on npm receives security fixes. Older versions are not patched; users should upgrade to the current release.

Version Supported
Latest (1.x) Yes
Older No

Reporting a Vulnerability

Do not open a public GitHub issue for security vulnerabilities.

Please report them privately via GitHub Security Advisories.

Include:

  • A description of the vulnerability and its potential impact
  • Steps to reproduce or a proof-of-concept (even a rough one helps)
  • Any suggested mitigation if you have one

You can expect:

  • Acknowledgement within 48 hours
  • A status update within 7 days (confirmed, needs more info, or not a vulnerability)
  • Coordinated disclosure — a fix will be released before the vulnerability is made public
  • Credit in the changelog if you would like it

Scope

The following are considered in scope:

  • Authentication bypass or token leakage in the HTTP server
  • Input validation bypasses that reach the Buda API with unintended data
  • Information disclosure of credentials, API keys, or account data to unauthorized callers
  • Dependency vulnerabilities in runtime dependencies (express, helmet, express-rate-limit, @modelcontextprotocol/sdk) that are exploitable in this server's context

The following are considered out of scope:

  • Prompt injection via API response content — the server returns data from the Buda.com API verbatim as JSON. An AI agent interpreting that data in unexpected ways is a risk at the agent/model layer, not in this server. There is no server-side mitigation for this class of issue.
  • confirmation_token='CONFIRM' bypass — this token is a UX safety guard to prevent accidental execution by AI agents, not a cryptographic access control. Any caller who holds a valid MCP_AUTH_TOKEN already has full account access by design.
  • Vulnerabilities in Buda.com's own API or infrastructure
  • Social engineering or phishing attacks
  • Bugs in MCP client applications (Claude Desktop, Cursor, etc.)

Security Model

stdio mode (recommended for personal use): The server process runs locally on your machine. Credentials are set as environment variables and never leave your system. There is no network exposure.

HTTP mode (self-hosted): The server exposes an HTTP endpoint protected by a bearer token (MCP_AUTH_TOKEN). This mode requires TLS termination (via Railway, Nginx, Caddy, etc.) when auth credentials are configured. Running over plain HTTP exposes your API key and secret to network interception.

The MCP_AUTH_TOKEN is the sole cryptographic boundary for the HTTP server. Anyone who holds a valid token has full access to all tools the server exposes, including fund movements. Treat it with the same care as your Buda API secret.

There aren't any published security advisories