# agent-tools MCP server

> Lets an AI client discover, search, and safety-scan x402 paid APIs, MCP servers, and A2A agents through an intent-based directory.

- Repository: https://github.com/AgentTools-Cloud/AgentToolsCollection
- Page: https://hostedmcp.org/servers/agenttools-cloud/agenttoolscollection
- Transport: stdio, remote
- Stars: 1
- Latest version: 0.5.2

## Install

### Claude Code

```sh
claude mcp add agenttoolscollection -- uvx agent-tools-mcp
```

### Cursor (`~/.cursor/mcp.json`)

```json
{
  "mcpServers": {
    "agenttoolscollection": {
      "command": "uvx",
      "args": [
        "agent-tools-mcp"
      ]
    }
  }
}
```

### VS Code (`.vscode/mcp.json`)

```json
{
  "servers": {
    "agenttoolscollection": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "agent-tools-mcp"
      ]
    }
  }
}
```

### Claude Desktop (`claude_desktop_config.json`)

```json
{
  "mcpServers": {
    "agenttoolscollection": {
      "command": "uvx",
      "args": [
        "agent-tools-mcp"
      ]
    }
  }
}
```

### Codex (`~/.codex/config.toml`)

```toml
[mcp_servers.agenttoolscollection]
command = "uvx"
args = ["agent-tools-mcp"]
```

### Gemini CLI (`~/.gemini/settings.json`)

```json
{
  "mcpServers": {
    "agenttoolscollection": {
      "command": "uvx",
      "args": [
        "agent-tools-mcp"
      ]
    }
  }
}
```

## FAQ

### Can I scan an MCP server for safety?

Yes, the scan_mcp_safety tool vets MCP servers for malware and prompt-injection lures using deterministic static rules and an LLM second opinion.

### How does the service search work?

The search tool accepts natural-language intent and ranks results using FTS5, popularity, and health metrics.