mcpdirectory.dev
9,129 servers · updated 23 Sept

Gateco MCP server

community Active last commit 19 d
{
  "mcpServers": {
    "gateco": {
      "command": "uvx",
      "args": [
        "gateco"
      ],
      "env": {
        "GATECO_API_KEY": "<gateco_api_key>"
      }
    }
  }
}
file .mcp.json (in your project root)install uvx gatecoversion 1.11.1

What it does

Gateco is an MCP server that sits between AI agents and vector stores, enforcing access policies before any document content reaches a model. It exposes tools for permission-gated retrieval, grounded answer synthesis, and access simulation, intended for developers building RAG pipelines that must respect organizational access controls.

  • Query a document connector on behalf of a specific user and receive only the chunks that user's access policies permit.
  • Generate a grounded answer from policy-filtered search results, with citations drawn only from content the principal is authorized to see.
  • Dry-run an access check to determine whether a given principal would be allowed to retrieve specific content before executing a live query.

Tools this server exposes

ToolDoes
gateco_retrieveExecutes a permission-aware retrieval against a connector using vector, keyword, hybrid, or grep search modes, returning only chunks the specified principal is
gateco_askGenerates a grounded answer synthesized from policy-filtered chunks, with citations limited to content the principal is permitted to see (requires Team+ plan).
gateco_check_accessPerforms a dry-run access simulation to preview whether a principal would be granted or denied access without executing a live retrieval (requires Growth+ plan)
gateco_list_connectorsLists available connectors along with their readiness levels.
gateco_list_principalsLists identity principals registered in the system.
gateco_resolve_principalLooks up a principal by email address or identity provider subject ID.

Before you enable it

An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.