PolicyLayer Registry MCP Server

The MCP server that vets MCP servers.

Connect your agent to the PolicyLayer registry and it can ask one question before installing or allowing any MCP server: is this safe? The answer is the published registry record: verified identity with the evidence, an A to F risk grade, probed auth posture and every tool's risk classification.

Endpoint: https://api.policylayer.com/mcp (Streamable HTTP)
Free single-server lookups, no key needed
32,500+ published servers and 515,000+ classified tools, kept current by continuous scanning
Docs: policylayer.com/registry/mcp

Connect

Claude Code

Terminal

claude mcp add --transport http policylayer https://api.policylayer.com/mcp

Claude (desktop / web)

Settings → Connectors → Add custom connector → URL https://api.policylayer.com/mcp

Cursor / Windsurf / VS Code

config.json

{ "mcpServers": {
  "policylayer": {
    "url": "https://api.policylayer.com/mcp"
} } }

Then tell your agent: "check any MCP server against PolicyLayer before installing it." Add it to your team's agent instructions and every install gets vetted.

Tools

Tool	What it returns	Tier
`check_mcp_server`	Full record for one server: identity + evidence, risk grade, posture, tools riskiest-first. Unknown servers are queued for scanning by the lookup itself.	Free
`search_registry`	Published servers matching a name, slug or package substring, with grade and verification on every match.	Free
`check_tool`	One tool's full classification: risk analysis and evidence, OWASP classes, parameters, recommended policy default.	Free
`get_change_events`	The ordered change feed: drift, posture flips, impostor flags. Cursor-based, with severity filters.	Licence

Example

Ask your agent "is the github MCP server safe to install?" and it calls:

config.json

{ "name": "check_mcp_server", "arguments": { "server": "github" } }

and gets back the published record (trimmed):

config.json

{
  "slug": "github",
  "riskGrade": "D",
  "identity": { "verification": "official", "confidence": "verified" },
  "posture": { "auth": "gated", "rateLimited": false },
  "toolCount": 86,
  "tools": [
    { "name": "delete_file", "category": "Destructive", "severity": "High", "recommendedPolicy": { "action": "deny" } }
  ]
}

Things to ask your agent

"Is the github MCP server safe to install?"
"We're adding the notion and linear MCP servers this week. Vet both and summarise the risks."
"Which of the MCP servers in this project can delete data or move money?"
"Should I allow the execute_sql tool on the postgres server? What policy does PolicyLayer recommend?"
"Find a verified MCP server for Slack and show me its risk grade."
"Check every server in my mcpServers config against PolicyLayer and flag anything below grade C."
With a licence key: "What changed in the servers we depend on since yesterday? Any posture flips or impostor flags?"

Adding a licence key

get_change_events needs a Registry Licence key (plr_...), self-serve at policylayer.com/registry/pricing.

Claude Code: append --header "Authorization: Bearer plr_..." to the command above
Cursor / Windsurf / VS Code: add "headers": { "Authorization": "Bearer plr_..." } inside the server entry
Claude desktop/web connectors can't send headers; free lookups only

Free and licensed

Everything about a single server is free and complete: the full record, every tool, the whole classification. A Registry Licence adds breadth across the catalogue: the change feed above, plus bulk snapshots, keyset paging and webhooks on the /v1 API.

What this repository is

The connect kit for a hosted service. The server runs at api.policylayer.com; there is no code to run from here. server.json is the listing published to the official MCP Registry as com.policylayer/registry. The MIT licence covers the contents of this repository only; the hosted service and the registry data are licensed separately.

PolicyLayer · Registry lookup · API reference · Licensing