One endpoint, https://neuronto.com/mcp, answers every client below. It searches this index and every other public ARD registry at once and fuses the rankings, so a single connector covers the federation rather than one catalogue.

No account, no key, no signup. Anonymous callers get 60 requests an hour; proving a domain raises that to 300, and a keyed search is not logged at all (see privacy).

Claude Code

Installs a skill, so asking for a tool in plain language searches ARD without you naming a registry. The plugin bundles the connector, so this is the whole setup.

Two commands, in Claude Code
/plugin marketplace add neuronto/ard-connectors
/plugin install neuronto-agent-finder@neuronto

Then: /agentfinder a tool that can read PDFs. The marketplace lists every public Agent Finder, not only this one. Source: neuronto/ard-connectors.

Editors and IDEs

Cursor

~/.cursor/mcp.json, or .cursor/mcp.json in a project
{
  "mcpServers": {
    "neuronto": {
      "url": "https://neuronto.com/mcp"
    }
  }
}

VS Code and GitHub Copilot

.vscode/mcp.json
{
  "servers": {
    "neuronto": {
      "type": "http",
      "url": "https://neuronto.com/mcp"
    }
  }
}

Or run MCP: Add Server from the command palette and choose HTTP.

Zed

Zed settings.json
{
  "context_servers": {
    "neuronto": {
      "url": "https://neuronto.com/mcp"
    }
  }
}

Or Settings, AI, MCP Servers, Add Server, Add Remote Server. Zed is deprecating its own MCP extensions in favour of the official MCP registry, where this server is already published.

Windsurf (Cascade)

~/.codeium/mcp_config.json
{
  "mcpServers": {
    "neuronto": {
      "serverUrl": "https://neuronto.com/mcp"
    }
  }
}

Cascade uses serverUrl for remote HTTP servers rather than url. Press refresh after adding it.

JetBrains AI Assistant

Settings, Tools, AI Assistant, MCP, New MCP Server, as JSON
{
  "mcpServers": {
    "neuronto": {
      "url": "https://neuronto.com/mcp"
    }
  }
}

Choose the Streamable HTTP transport. Works across IntelliJ IDEA, PyCharm, WebStorm, Rider and the rest of the family.

Continue

.continue/mcpServers/neuronto.yaml
name: Neuronto
version: 0.0.1
schema: v1
mcpServers:
  - name: Neuronto
    type: streamable-http
    url: https://neuronto.com/mcp

MCP tools are available in agent mode.

Terminals and CLIs

Claude Desktop

claude_desktop_config.json
{
  "mcpServers": {
    "neuronto": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://neuronto.com/mcp"
      ]
    }
  }
}

Settings, Developer, Edit Config. Restart Claude after saving.

Gemini CLI

~/.gemini/settings.json
{
  "mcpServers": {
    "neuronto": {
      "httpUrl": "https://neuronto.com/mcp"
    }
  }
}

Codex CLI

MCP server config
{
  "mcpServers": {
    "neuronto": {
      "type": "http",
      "url": "https://neuronto.com/mcp"
    }
  }
}

Warp

MCP servers, + Add, paste JSON
{
  "mcpServers": {
    "neuronto": {
      "url": "https://neuronto.com/mcp"
    }
  }
}

Goose

One-click extension link
goose://extension?url=https%3A%2F%2Fneuronto.com%2Fmcp&type=streamable_http&id=neuronto&name=Neuronto&description=Find%20MCP%20servers%2C%20skills%2C%20agents%20and%20APIs%20across%20every%20public%20ARD%20registry

Open that link with Goose installed and it adds the extension. Or add a remote extension by hand with type streamable_http.

Agent platforms

Anything that can call an MCP server or an HTTP endpoint can use this index. In n8n, add the MCP Client Tool node and point it at https://neuronto.com/mcp with HTTP Streamable transport. In Cline and Roo Code, use Configure MCP Servers and add a remote server with the same URL. Dify, LangFlow and similar tool-calling platforms take it as either an MCP server or a plain REST call.

Anything else

The index answers on three interfaces and they share one search path, so they cannot disagree about what it holds.

REST, the ARD interface
curl -s -X POST https://neuronto.com/search \
  -H "Content-Type: application/json" \
  -d '{"query":{"text":"read a PDF and extract tables"}}'
A2A, for agent-to-agent clients
curl -s -X POST https://neuronto.com/a2a \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"message/send","params":{"message":{"messageId":"1","role":"user","parts":[{"kind":"text","text":"read a PDF"}]}}}'

The A2A agent card is at /.well-known/agent-card.json. The MCP endpoint is POST only and answers 405 to GET: there is no server-initiated stream, because every tool answers inside the request that asked. Full reference at /api-docs, and /agents.md if you are an agent reading this yourself.

One page per client

Each of these covers that client on its own, including the quirk that makes a config copied from somewhere else fail in it.

What you get back

Ranked matches with the endpoint to connect to, which registries carried each one, and what we verified by fetching it: whether it answered, and the tools its own tools/list returned. The score is semantic relevance only. It is not a trust, safety or quality rating and must not be shown to anyone as one.