Skip to main content
hc
Submit
commandsSource-backedReview first Safety ✓ Privacy ✓

/mcp-auth-audit - MCP Authorization Audit Command

Slash command for auditing a remote MCP server authorization setup, including protected resource metadata, authorization server discovery, OAuth resource indicators, token audience validation, scopes, and token passthrough risk.

by JSONbored·added 2026-06-05·
Claude Code
HarnessClaude Code
Invocation:/mcp-auth-audit <server-url>
Review first — review before installing

Open the source and read safety notes before installing.

Safety notes

  • The command may ask the assistant to inspect authentication metadata and token-flow evidence; do not paste production bearer tokens into the prompt.
  • Treat missing resource metadata, wrong-audience token acceptance, or token passthrough as blocking findings.

Privacy notes

  • Metadata and audit output can expose identity providers, scopes, tenant identifiers, redirect URLs, and private server URLs.
  • Redact token claims and account identifiers before sharing command output publicly.

Prerequisites

  • Remote MCP server URL and expected canonical resource URI.
  • Approved staging credentials or metadata-only access for authorization review.
  • Permission to inspect HTTP responses, authorization metadata, and token-flow evidence.

Schema details

Install type
cli
Troubleshooting
No
Runtime and command metadata
Command syntax
/mcp-auth-audit <server-url>
Full copyable content
/mcp-auth-audit <server-url>

About this resource

The /mcp-auth-audit command turns the MCP authorization specification into a repeatable review workflow for remote servers. It is useful before adding a new MCP endpoint to Claude Code, Cursor, a gateway, or an internal agent platform.

Usage

/mcp-auth-audit

What it does

  1. Captures the unauthenticated 401 challenge and checks whether it advertises protected resource metadata.
  2. Fetches resource metadata and authorization server metadata.
  3. Checks that the client flow includes the MCP server as the resource.
  4. Reviews audience binding and wrong-audience rejection evidence.
  5. Reviews scopes, redirect handling, dynamic client registration, and error behavior.
  6. Flags token passthrough or acceptance of tokens issued for another resource.
  7. Produces a public-safe summary plus private evidence handling notes.

Output format

  • Server reviewed
  • Metadata discovery result
  • Authorization server result
  • Resource indicator result
  • Audience validation result
  • Scope and redirect notes
  • Blockers
  • Decision

References

#mcp#oauth#slash-command#security-audit#authorization

Source citations

Signals

Loading live community signals…

More like this, weekly

A short, calm digest of reviewed Claude resources. Unsubscribe any time.