GitLab 18.3: Expanding AI orchestration in software engineering

What this post added

This post details the expansion of GitLab's AI orchestration capabilities in version 18.3. Key contributions include the introduction of a Knowledge Graph for real-time code indexing and faster, more contextual code searches. A new orchestration layer is added to the control plane, enabling agents to subscribe to SDLC events, a multi-agent flow engine, and exposing GitLab tools via MCP and standard protocols. The user experience is enhanced with native agents and agent flows across the SDLC, supporting async tasks, @ mentions, and custom agents. Two new automated development flows are introduced: 'Issue to MR Flow' for generating code from issues and 'Convert CI File Flow' for migrating CI/CD configurations (initially supporting Jenkins). Integrations with third-party AI models (Claude Code, Codex, Amazon Q, Google Gemini, opencode) are enabled via @ mentions in issues/MRs, requiring BYOK. The MCP server is introduced for universal AI integration with tools like Cursor. Agentic Chat support is extended to Visual Studio IDE. Expanded AI model support includes OpenAI GPT models and Anthropic's Claude 4 for self-hosted deployments.