MCP Server should be available without cloud-connected AI Gateway for self-hosted/air-gapped customers".

Everyone can contribute. Help move this issue forward while earning points, leveling up and collecting rewards.

Summary

The GitLab MCP Server currently requires Duo Core to be enabled, which in turn requires a connection to GitLab's cloud-connected AI Gateway. This creates a significant gap for customers who cannot or choose not to use cloud-connected services.

Problem

  1. MCP Server requires Duo Core: The MCP Server feature is gated behind Duo Core enablement
  2. Duo Core requires cloud connectivity: Duo Core requires connection to GitLab's cloud AI Gateway
  3. Self-hosted models require Enterprise: Customers with Core/Pro licenses who want to use self-hosted models (to avoid cloud connectivity) must upgrade to Duo Enterprise to access self-hosted model configurations

This creates barriers for:

  • Air-gapped customers: Organizations in highly regulated industries (defense, government, healthcare, finance) that cannot connect to external cloud services
  • Data sovereignty requirements: Customers with strict data residency requirements who cannot send data to cloud AI services
  • Core/Pro customers: Users who have Duo Core or Pro but would need to upgrade to Enterprise solely to access self-hosted models as a workaround

Current Behavior

Expected Behavior

MCP Server should be usable in self-hosted GitLab deployments without requiring:

  1. Connection to GitLab's cloud AI Gateway, OR
  2. An upgrade path that doesn't require Enterprise tier for basic MCP functionality

Proposal

Consider one or more of the following approaches:

  1. Decouple MCP Server from Duo Core: Allow MCP Server to function independently as an API integration layer, separate from AI features
  2. Enable MCP Server for self-hosted configurations: Allow customers using self-hosted AI Gateway to access MCP Server without cloud connectivity requirements
  3. Tier adjustment: Make self-hosted model support available at lower tiers (Core/Pro) for customers who cannot use cloud-connected services

Use Cases

Air-gapped Government Contractor

A defense contractor using GitLab Self-Managed in an air-gapped environment wants to use MCP Server with their own AI coding assistants but cannot enable Duo Core due to the cloud connectivity requirement.

Regulated Financial Institution

A bank with strict data governance policies wants to leverage MCP Server for developer productivity but cannot allow any data to flow to external cloud services.

Self-Hosted AI Customer

A customer already running self-hosted models via vLLM wants to use MCP Server but is blocked because they have Duo Pro (not Enterprise) and cannot enable Duo Core without cloud connectivity.

Related Issues

Edited by Dan Rabinovitz