# Gateco MCP server

> Enforces organizational access policies during RAG retrieval to prevent AI models from accessing unauthorized sensitive data.

- Repository: https://github.com/fortisil/gateco-sdk-python
- Page: https://hostedmcp.org/servers/fortisil/gateco-sdk-python
- Transport: stdio
- Stars: 0
- License: MIT
- Latest version: 1.13.0

## Install

### Claude Code

```sh
claude mcp add --env GATECO_API_KEY=<YOUR_GATECO_API_KEY> gateco-sdk-python -- uvx gateco
```

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

```json
{
  "mcpServers": {
    "gateco-sdk-python": {
      "command": "uvx",
      "args": [
        "gateco"
      ],
      "env": {
        "GATECO_API_KEY": "<YOUR_GATECO_API_KEY>"
      }
    }
  }
}
```

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

```json
{
  "servers": {
    "gateco-sdk-python": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "gateco"
      ],
      "env": {
        "GATECO_API_KEY": "<YOUR_GATECO_API_KEY>"
      }
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "gateco-sdk-python": {
      "command": "uvx",
      "args": [
        "gateco"
      ],
      "env": {
        "GATECO_API_KEY": "<YOUR_GATECO_API_KEY>"
      }
    }
  }
}
```

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

```toml
[mcp_servers.gateco-sdk-python]
command = "uvx"
args = ["gateco"]

[mcp_servers.gateco-sdk-python.env]
GATECO_API_KEY = "<YOUR_GATECO_API_KEY>"
```

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

```json
{
  "mcpServers": {
    "gateco-sdk-python": {
      "command": "uvx",
      "args": [
        "gateco"
      ],
      "env": {
        "GATECO_API_KEY": "<YOUR_GATECO_API_KEY>"
      }
    }
  }
}
```

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `GATECO_API_KEY` | yes | yes | Gateco API key WITH THE retrieve SCOPE (Settings > API Keys > Create, tick retrieve). Every plan. Six of seven tools run on it; gateco_check_access needs `gateco login`. |
| `GATECO_BASE_URL` | no | no | Gateco API base URL. Defaults to https://api.gateco.ai; set only for self-hosted or staging deployments. |

## Tools (7)

- `gateco_retrieve`
- `gateco_ask`
- `gateco_check_access`
- `gateco_list_connectors`
- `gateco_list_principals`
- `gateco_list_groups`
- `gateco_resolve_principal`

## FAQ

### How does Gateco prevent data leaks in RAG pipelines?

Gateco sits between your AI agent and your vector store, evaluating every retrieval against access policies before content reaches the model.

### Can I use Gateco with existing vector stores?

Yes, Gateco acts as a permission-aware layer for retrieval, supporting various search modes like vector, keyword, hybrid, and grep.