mcp-shield

mcp-shield is a CLI security scanner that inspects installed MCP server configurations/tools and reports potential prompt-injection/tool-poisoning style vulnerabilities, suspected data-exfiltration channels, tool shadowing/behavior modification, and cross-origin escalation risks. It can optionally use an Anthropic Claude API key to enhance analysis.

Evaluated Mar 30, 2026 (21d ago)
Repo ↗ Security mcp security cli prompt-injection tool-poisoning exfiltration-detection static-analysis anthropic typescript
⚙ Agent Friendliness
32
/ 100
Can an agent use this?
🔒 Security
38
/ 100
Is it safe for agents?
⚡ Reliability
29
/ 100
Does it work consistently?

Score Breakdown

⚙ Agent Friendliness

MCP Quality
20
Documentation
75
Error Messages
0
Auth Simplicity
85
Rate Limits
0

🔒 Security

TLS Enforcement
40
Auth Strength
35
Scope Granularity
20
Dep. Hygiene
55
Secret Handling
45

Security posture is mixed based on README evidence only: TLS enforcement, secret-handling practices, and error-output handling are not documented. The optional Anthropic API key increases the chance of accidental exposure if the tool logs requests/headers (not stated). The tool is designed to detect risky patterns (including exfiltration and hidden instructions) but this is not the same as guaranteeing that it will not leak data during scanning/AI analysis. Dependency hygiene is unknown; listed dependencies include common Node libraries but no vulnerability/CVE status is provided.

⚡ Reliability

Uptime/SLA
0
Version Stability
55
Breaking Changes
30
Error Recovery
30
AF Security Reliability

Best When

You are integrating new or updated MCP servers and want a fast static risk review before enabling them in an agent/LLM workflow.

Avoid When

You need verifiable, deterministic guarantees of safety (scanner outputs are best-effort) or you cannot risk any external API calls (when using optional Claude-based analysis).

Use Cases

  • Scan MCP server tool descriptions before integrating them into an LLM client
  • Run periodic security audits of local MCP configuration files
  • Assess risk after MCP server updates
  • Validate MCP server behavior around messaging/cross-origin access patterns
  • Help identify suspicious tool parameters that could be used for exfiltration

Not For

  • Not intended as a full security guarantee or runtime protection (it does static/config analysis)
  • Not a replacement for code review or sandboxed execution testing of MCP servers
  • Not suitable for environments requiring strict offline operation if AI-enhanced mode calls external APIs

Interface

REST API
No
GraphQL
No
gRPC
No
MCP Server
No
SDK
No
Webhooks
No

Authentication

Methods: Optional --claude-api-key (Anthropic Claude API key)
OAuth: No Scopes: No

Authentication is only relevant for optional AI-enhanced analysis via Anthropic. No OAuth flow or scopes are described.

Pricing

Free tier: No
Requires CC: No

No pricing details provided for the CLI itself; costs may arise from calling Anthropic when enabled.

Agent Metadata

Pagination
none
Idempotent
False
Retry Guidance
Not documented

Known Gotchas

  • This is a local CLI scanner; there is no documented programmatic API for agents beyond invoking the command.
  • AI-enhanced mode depends on external Anthropic API credentials and may introduce non-determinism/cost.
  • Risk patterns rely on scanning tool descriptions/config; false positives/negatives are possible and not quantified in README.

Alternatives

Full Evaluation Report

Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for mcp-shield.

AI-powered analysis · PDF + markdown · Delivered within 30 minutes

$99

Package Brief

Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.

Delivered within 10 minutes

$3

Score Monitoring

Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.

Continuous monitoring

$3/mo

Scores are editorial opinions as of 2026-03-30.

8642
Packages Evaluated
17761
Need Evaluation
586
Need Re-evaluation
Community Powered