Feature Comparison
Honest side-by-side. We highlight where skills.sh wins on scale and where we win on verification depth.
| Feature | My MCP Shelf | skills.sh |
|---|---|---|
| Security grades per listing Both grade listings; we add execution testing and staleness. | ✓ Yes | ✓ Yes |
| Multiple audit sources (Snyk, Socket, etc.) skills.sh uses Gen Agent Trust Hub + Socket + Snyk; we run our own unified pipeline plus external audits. | ✓ Yes | ✓ Yes |
| Execution test (does the server handshake?) | ✓ Yes | ✗ No |
| Staleness pulse (commit velocity) | ✓ Yes | ~ Partial |
| Green hosting / carbon checks | ✓ Yes | ✗ No |
| Agent-specific pages skills.sh has ~20 agent pages; we generate them programmatically from live data. | ✓ Yes | ✓ Yes |
| Public JSON API | ✓ Yes | ~ Partial |
| Installable MCP server for agents | ✓ Yes | ✗ No |
| llms.txt for AI agent discovery | ✓ Yes | ✗ No |
| Embedddable live status badges | ✓ Yes | ✗ No |
| Server install counts (CLI telemetry) | ✓ Yes | ✗ No |
| Remote server auth & uptime monitoring | ✓ Yes | ✗ No |
| Focus area We verify the servers agents use; they verify the skills agents wear. | MCP servers + skills | Agent skills + harness components |
Common Questions
Are you trying to replace skills.sh?
No. skills.sh is excellent for agent skills and harness components. We are the directory for the MCP servers underneath those skills. Many developers use both.
Whose security grades are more reliable?
Both use multiple sources. Our differentiator is the execution-test harness: we actually spin up verified servers and confirm they complete an MCP handshake. That is ground truth, not a proxy.
Do you list skills too?
Yes — we have Agent Skills and Cursor Skills sections. But our primary depth is MCP servers, where we run the unified recheck pipeline (tasks 13/14/19).
Need servers, not just skills?
Browse MCP servers with execution-tested badges, security grades, and per-agent install commands.