GitLab Duo Agent Platform: general availability of agents across the delivery lifecycle
2026-01-15 · Publisher : GitLab · Domain : Cloud & DevOps
On 15 January 2026, GitLab declares the Duo Agent Platform generally available to all Premium and Ultimate customers: agentic chat, Planner and Security Analyst agents, flows to fix a CI/CD pipeline, and external agents via MCP.
What changes
- The GitLab Duo Agent Platform is generally available on 15 January 2026.
- Planner and Security Analyst agents, and flows to fix or convert CI/CD pipelines.
- Support for external agents (Claude Code, Codex CLI) and MCP-compatible systems.
- GitLab credits included: $12 per user per month in Premium, $24 in Ultimate.
### Background On 15 January 2026, GitLab announces general availability of the GitLab Duo Agent Platform, in version 18.8 of its DevSecOps platform. It brings AI agents that draw on the full project context (issues, merge requests, pipelines, security findings) and on the organization's governance rules. ### Key facts - Agentic Chat: contextual assistance in the web interface and IDEs, with multi-step reasoning, including for diagnosing CI/CD problems. - Two foundational agents: Planner (structure and prioritize work) and Security Analyst (review vulnerabilities and explain their impact). - Agentic flows provided: issue to merge request, conversion to GitLab CI/CD, fixing a failing pipeline, code review and development in the IDE. - Custom agents through a catalogue, and external agents such as Claude Code (Anthropic) or Codex CLI (OpenAI); connection to MCP-compatible systems. - Available on GitLab.com and self-managed for Premium and Ultimate; usage consumes GitLab credits, $12 per user per month included in Premium, $24 in Ultimate. ### Why it matters GitLab places the agent at the heart of the delivery platform, with DevOps-specific tasks: fixing a pipeline, converting CI scripts, analysing vulnerabilities. The credit-based model and openness to external agents via MCP show the market organizing around agent orchestration rather than a single assistant. For platform teams, governance will be the question: agent rights, credit costs and auditability. ### Takeaway GitLab installs agents in the pipeline itself, with credit-based pricing and openness to MCP and third-party agents.
Links
Tags : Agents, CI/CD, MCP, DevOps, IA