MCP Risk

Directory / server

ai.foura/mcp

This directory entry reports an automated security scan of ai.foura/mcp, scanned 2026-07-28. Trust grade B rates provenance, transport, known vulnerabilities and instruction integrity; capability risk is reported separately, not folded into the grade.

registry.modelcontextprotocol.io, scanned 2026-07-28, tool surface via remote-tools-list

B 85/100 permalink

Automated checks

StatusCheckPolicyResult
PASS Listed on the official MCP registry §1.1
Listed on registry.modelcontextprotocol.io as ai.foura/mcp. 1 finding
  • Registry entry: ai.foura/mcp ↗
PASS Source repository is public and maintained §1.2
Active repository with a license. 5 findings
  • Repository ↗
  • Last push: 4 days ago
  • Stars: 4
  • License: MIT
  • Security policy (SECURITY.md): absent
PASS Published package is consistent and attested §1.3
Package is consistent with its source and carries provenance attestation. 6 findings
  • Package: @fouradata/mcp@0.5.0 ↗
  • Ecosystem: npm
  • Age: 76 days
  • Weekly downloads: unknown
  • Repo field matches source repo: true
  • Build-provenance attestation: present
PASS Tool surface is inspectable and proportionate §2.1–§2.4
4 tool(s) enumerated. Capabilities detected: network-egress. Reported as capability risk (§2.4); whether each is essential to the server's stated purpose is §2.3 and remains a manual judgement. 4 findings
  • Tools exposed: 4
  • foura_single: network-egress
  • foura_proxy: network-egress
  • Detected capabilities: network-egress
PASS HTTPS-only endpoint §3.1
Endpoint uses HTTPS. 1 finding
  • Endpoint: https://mcp.foura.ai/mcp
FAIL Authentication required §3.2

Server answers anonymous requests (HTTP 200) while exposing non-trivial or unverifiable capabilities - fail per §3.2.
UNVERIFIABLE OAuth protected-resource metadata published §3.3

Metadata endpoint unreachable: Too many subrequests by single Worker invocation. To configure this limit, refer to https://developers.cloudflare.com/workers/wrangler/configuration/#limits
PASS No known vulnerabilities (OSV.dev) §4.1

No advisories on record for @fouradata/mcp @0.5.0 (npm).
PASS Server-supplied instructions free of poisoning indicators §5.1–§5.4

No poisoning indicators found across 14 server-supplied text item(s).

Score history

DateGradeScore
2026-07-28B85/100
2026-07-27B84/100
2026-07-27B84/100
2026-07-26B84/100
2026-07-26B84/100
2026-07-26B84/100
2026-07-25F53/100
2026-07-25F50/100
2026-07-25F53/100
2026-07-25F53/100
2026-07-24F53/100
2026-07-24F53/100
2026-07-24F53/100
2026-07-13C65/100
What this grade covers
  • Provenance & supply-chain integrity (§1)
  • Tool surface & capability risk (§2)
  • Transport & authentication (§3)
  • Known CVEs via OSV.dev (§4)
  • Instruction/tool poisoning patterns (§5)

Limitations: pattern-based only (no semantic injection detection), static npm analysis (live server instructions not retrieved for package-only scans), single-server scope (no cross-server toxic flow analysis). See the full policy.

Methodology: MCP Server Security Policy. Unfamiliar term? See the glossary.