Skip to content

Toolcard MCP server

Which MCP servers are Toolcard's own, from the official registry and from asking its own domain, whether they answer, and what they offer. A program checked each of these, nobody copied them from a product page.

read 2026-10-28

← Back to Toolcard

Whose server it is
  • Registry namecom.agentghostedme/toolcardverified in the MCP registry

The registry name sits under agentghostedme.com. The registry only lets a publisher use that namespace after proving control of the domain through DNS or HTTP.

When we asked itchecked 2026-10-28
  • Stateanswerschecked 2026-10-28
  • Tools4
  • Server calls itselftoolcard
  • Server version0.1.8
  • Protocol version2025-06-18

On 2026-10-28 the server answered initialize and tools/list without a sign-in. First answer in 486 ms.

Should an agent connect itConnect it
  • Its tools do every use case we checked: catch vague tool names and thin descriptions before release; compare a tool list for duplicate names and overlapping descriptions; record whether an agent called or ignored a tool.

Once connected

  • For "catch vague tool names and thin descriptions before release", the likely first call is check_tool.
  • For "compare a tool list for duplicate names and overlapping descriptions", the likely first call is check_server.
  • For "record whether an agent called or ignored a tool", the likely first call is report_outcome.

Said by rules from the measurements above. The likely first call is a model's pick from the tool list; read its description before relying on it.

Measured dailydaily checks: 7, the first on 2026-10-28
  • INITE MCP score100 of 100answers 100 · does the job 100 · definitions 100 · context 100 · connects 100
  • What an agent can do with it3 of 3 use cases
  • Answered100% of days7 of 7 daily checks
  • Median first answer500 ms
  • Space in an agent's context≈ 530 tokensthe definitions of all its tools (4), which an agent carries on every turn
  • Tool definitionsA · 100 of 100
  • Typed results0 of 4 toolsdeclared outputSchema: results come typed, read once
  • Every tool passes all six checks.

What an agent can do with it

  • can do · catch vague tool names and thin descriptions before release
  • can do · compare a tool list for duplicate names and overlapping descriptions
  • can do · record whether an agent called or ignored a tool

Each of the product's use cases, put to Jev against this server's tool list: can an agent do it with these tools?

Worth a look

None of the five warning signs in these definitions.

A program checks the server once a day: initialize and tools/list, no model. Tokens are the definitions' length at about four characters a token. The grade averages six checks of every tool's name, description, schema and annotations; it says how well the tools are described, not whether they work. The score weighs what was measured and nothing else: answering (25%), doing the product's job (25%), definitions (20%), context (10%), use (10%) and how readily a client connects (10%). A part that was not measured is left out, and a hidden instruction caps the score at 40. The state of vendors' MCP servers: method and numbers →

Tools it offers4 tools · read 2026-10-28
  • check_tool

    Check one MCP tool name, description, and input schema against Toolcard cited rules. Returns findings. A finding with an answer page includes resource, a toolcard://answers URI to pass to resources/read. Does not store the tool.

  • check_server

    Check every tool in a tools/list result or a server card. Returns per-tool findings plus duplicate-name, overlapping-descriptions, and too-many-tools. Those three omit resource. Does not store the tools.

  • report_outcome

    Record whether an agent called or ignored a tool. Stores the name, a SHA-256 of the description, and the result. Does not store the raw description.

  • outcome_summary

    Return the anonymous tally of called versus ignored reports on this server. Not a representative sample.

How to connect

Add the server to your MCP client config. Claude Desktop, Cursor and most other clients read this shape:

{
  "mcpServers": {
    "toolcard": {
      "type": "http",
      "url": "https://agentghostedme.com/mcp"
    }
  }
}

Add it

Claude Code, in a terminal

claude mcp add --transport http agentghostedme https://agentghostedme.com/mcp

Add to CursorAdd to VS Code

claude.ai, ChatGPT and other clients: add https://agentghostedme.com/mcp as a custom connector in the settings.

OpenAPI spec

7 paths, read 2026-10-28. Open the spec ↗

Each line is something a program found when it went looking: the registry entry, the server's own answer, a spec that parses. A missing line means we did not find it, not that it does not exist.