Guard · MCP Scanner

Inspect MCP code and declared tool reach.

Analyse a public GitHub MCP repository or pasted configuration and tool text. Guard reports static code, permission and prompt signals with the evidence it matched.

Public repositoryPasted sourcePasted configPrompt-injection signals
Try:

Guard reads the package — manifest, prompts, source, dependencies — and reports what it observes against 11 check families. Public GitHub repositories are fetched live; for private code, paste only the source or manifest you are permitted to analyse. ZIP upload is not enabled in this release.

Guard reports observed security signals. A low-risk result is not a guarantee of safety, and no result should replace your own review of code you intend to run.

Guard performs bounded static text analysis. It does not complete an authenticated MCP handshake, execute tools, fetch arbitrary MCP endpoints, enumerate a private server or prove actual runtime reach. Paste the relevant configuration or tool listing directly when it is not in a public GitHub repository.

What it inspects

Six signal areas in the content Guard can collect.

Tool declarations

Wildcard capability grants and requests for automatic approval found in supplied text.

Network signals

Outbound calls, hard-coded IP addresses and patterns that post local data externally.

Credential signals

Embedded secrets, environment sweeps and reads of common credential files.

Prompt surface

Descriptions or instructions that try to override prior guidance or reveal context and keys.

Package behavior

Install hooks, obfuscation, vendored minified code and embedded binary indicators.

Permission mismatch

Observed static signals that appear broader than a supplied permission declaration.