MCP Observatory: one read only walk over every https endpoint in the official registry
Public record of one read only walk over the official MCP registry: who answered, tools, who pays.
Should I use this
Quality & Safety
Findings (1)
- LOWin mcp_observatory_disclosure_guide
Based on automated analysis of tool definitions and protocol compliance.
Context Cost
This is the approximate number of tokens consumed each time the server's tools are loaded into a model's context. Higher counts reduce the attention available for other tasks.
Install
One-Click Install
Add this to your `claude_desktop_config.json` file:
{
"mcpServers": {
"hs-mcp-observatory": {
"url": "https://observatory.horizonshield.dev/mcp"
}
}
}Remote endpoints
https://observatory.horizonshield.dev/mcpstreamable-httpWhat it can do
Tool inventory
Tools (6)
🟢mcp_observatory_state
What was measured, when, over what population, and what the measurement cannot tell you. Read this before quoting any number from this server.
Input Schema
{
"type": "object",
"properties": {}
}⚪mcp_observatory_summary
The counts, as published and as corrected. How many addresses answered, how many tools they exposed, how many stated who compensates their operator.
Input Schema
{
"type": "object",
"properties": {}
}⚪mcp_observatory_lookup(address, host)
What happened when we contacted one address, or an aggregate for one host. Returns the record hash and the procedure to recompute it. There is no tool here that lists addresses; you must already hold the one you are asking about.
Input Schema
{
"type": "object",
"properties": {
"address": {
"type": "string",
"description": "The full declared endpoint, e.g. https://example.com/mcp"
},
"host": {
"type": "string",
"description": "A host, e.g. example.com. Returns counts for that host, not a roster."
}
}
}⚪mcp_observatory_disclosure_guide
How the 152 servers that stated who compensates their operator actually did it - the field names observed in the wild, with counts. Not a specification we invented.
Input Schema
{
"type": "object",
"properties": {}
}🟡mcp_observatory_measure_now(address)
Contact one https address now and report what we see: whether it speaks MCP, whether it carries an agent card, and whether that card names who compensates the operator. Read only; no tool is called. robots.txt is read first and honoured. This does NOT change the published record - the next full walk does that. Rate limited, because it contacts somebody else's server on your behalf.
Input Schema
{
"type": "object",
"properties": {
"address": {
"type": "string",
"description": "https endpoint, e.g. https://example.com/mcp"
}
},
"required": [
"address"
]
}⚪mcp_observatory_method
How the walk was run, what we got wrong and corrected, and the four rules the report holds itself to.
Input Schema
{
"type": "object",
"properties": {}
}Community
Evidence