Loading…
A curated registry of Model Context Protocol servers, each with an automated security grade from our tool-surface scanner.
Context for the grades below — how inspection, tool poisoning, and the grade-vs-gate split work. Directory letters are Andrax scanner grades (mcpgrade-1.4.0), not Guard runtime scores.
Soft CTA: a grade finds risk on a snapshot; Guard enforces allow/deny/audit on live MCP traffic (mcpgrade-2.0.0). Do not equate the two scores.
Validates and generates EU e-invoices (UBL 2.1 / EN 16931) through a public, no-signup hosted MCP server.
Structured Japan sports schedules, fixtures, broadcasts, and live updates for AI agents via the Spocal API.
Exact VRAM and KV-cache fit calculations for LLMs on GPUs, multi-GPU rigs, and Macs — read-only MCP access.
Inspects webhook health, investigates delivery failures, configures sources, and replays missed events for Relin's webhook reliability platform.
Run an MCP server? Paste its remote endpoint — we'll scan it (read-only) and publish a free, reproducible security grade. Genuine MCP servers get listed instantly.