Developers
Squatch MCP server
Let your AI assistant look up an MCP server before it connects to one. Endpoint:
https://squatchmcp.com/mcpConnect
Squatch speaks Streamable HTTP and supports MCP protocol versions 2026-07-28 back to 2025-03-26. Most clients accept a configuration like this:
{
"mcpServers": {
"squatch": {
"type": "http",
"url": "https://squatchmcp.com/mcp"
}
}
}All 11 tools are read-only and work without an account (anonymous clients share a per-IP hourly limit). With an API key from your dashboard, calls count against your plan's daily API quota instead:
{
"mcpServers": {
"squatch": {
"type": "http",
"url": "https://squatchmcp.com/mcp",
"headers": {
"Authorization": "Bearer ${SQUATCH_API_KEY}"
}
}
}
}Tools
| Tool | Area |
|---|---|
| search_mcp_serversSearch the public catalog of MCP servers by name, publisher, tool, capability or category. Returns verification flags, tool effects and open-finding severity for each match. | Discovery |
| find_servers_by_capabilityFind MCP servers whose tools have a given capability class (e.g. FILE_ACCESS, PAYMENT) and/or effect class (READ, WRITE, DELETE, ADMIN, EXECUTE). | Discovery |
| get_mcp_serverProfile of one MCP server: publisher, verification flags, trust dimension statuses, latest version, transports, tool effects and capabilities. | Discovery |
| get_trust_profileAll nine trust dimensions for a server (registry namespace, publisher identity, organization, domain, repository, source provenance, security scan, privacy transparency, monitoring) with status, evidence and evidence source. There is no overall score. | Trust |
| get_server_toolsTools a server advertises, with Squatch's effect and capability classification, risk notes, declared annotations (unverified hints) and parameter names. | Inventory |
| get_security_findingsPublic findings from Squatch's metadata scanner (hidden content, prompt-injection patterns, risky tool design, authentication and transport issues) with status and publisher responses. | Security |
| get_privacy_disclosureThe latest publisher-declared privacy disclosure (data categories, retention, AI training, subprocessors, regions). Publisher statements are labelled as declared, not audited. | Privacy |
| get_mcbomMCP Bill of Materials (schema_version 1.0): a deterministic inventory of tools, resources, prompts, transports, auth, external integrations, data handling, findings and trust statuses. | Inventory |
| compare_mcp_serversSide-by-side facts for two to four servers: verification, trust dimension statuses, tool effects and capabilities, open findings and privacy completeness. | Trust |
| check_server_changesWhat changed on a server recently: tools added/removed/changed, permission expansions, new destructive capabilities, endpoint/auth/repository changes, verification changes. Default window 30 days. | Monitoring |
| assess_tool_capabilityExplain what one tool on a server appears able to do: Squatch's effect and capability classification, risk notes, whether its declared readOnly/destructive hints match the observed effect, and related findings. | Security |
Untrusted content
Tool results quote names and descriptions from third-party servers. Squatch wraps that text in fields marked untrusted and escapes hidden characters, so your assistant can show it to you without treating it as instructions. A well-behaved client should still ask you before acting on anything a looked-up server says.