GitLab Duo Agent Platform | GitLab Docs
- Tier: Premium, Ultimate
- Offering: GitLab.com, GitLab Self-Managed, GitLab Dedicated
The GitLab Duo Agent Platform is an AI-native solution that embeds multiple intelligent assistants (“agents”) throughout the software development lifecycle.
- Instead of following a linear workflow, collaborate asynchronously with AI agents.
- Delegate routine tasks, from code refactoring and security scans to research, to specialized AI agents.
To get started, see Get started with the GitLab Duo Agent Platform.
Prerequisites
To use the Agent Platform:
- Have GitLab Duo turned on.
- If you do not have GitLab Duo Pro or Enterprise, have GitLab Duo Core turned on for the top-level group or instance.
- Depending on your GitLab version:
- In GitLab 18.8 and later, have the Agent Platform turned on.
- In GitLab 18.7 and earlier, have beta and experimental features turned on.
- For GitLab Self-Managed, configure your instance.
- For GitLab Duo Self-Hosted, install the AI Gateway with the Agent Platform service.
To use the Agent Platform in your local environment:
- Install an editor extension and authenticate with GitLab.
- Have a project in a group namespace.
- Have the Developer, Maintainer, or Owner role.
Generally available features
These features are generally available and consume GitLab Credits when used.
Features available on the Free tier for GitLab.com customers require the purchase of GitLab Credits.
| Feature | Free | Premium | Ultimate |
|---|---|---|---|
| GitLab Duo Chat (agentic) Answer complex questions and autonomously create and edit files. | |||
| Code Suggestions Get AI-powered suggestions as you write code. | |||
| Custom agents Build team-specific agents for your unique development requirements. | |||
| External agents Securely connect third-party integrations and tools to extend Agent Platform capabilities. | No | ||
| Planner Agent Plan, prioritize, and track work. | |||
| Developer Flow Convert issues into merge requests. | |||
| Code Review Flow Automate code review tasks and enforce coding standards across your team. | |||
| Convert to GitLab CI/CD Flow Convert legacy CI/CD pipelines to the GitLab CI/CD format. | |||
| Fix CI/CD Pipeline Flow Diagnose and automatically fix failing CI/CD pipelines. | |||
| Software Development Flow Create a full, multi-step plan before executing it. | |||
| MCP clients Access GitLab resources and tools from any MCP-compatible AI client or IDE extension. 1 | |||
| SAST False Positive Detection Flow Automatically identify and filter out false positives in SAST security scans. | No | No | |
| Security Analyst Agent Automate repetitive security tasks: Triage issues, analyze vulnerabilities, and generate fixes. | No | No |
Footnotes:
- MCP clients do not consume credits directly. However, any Agent Platform usage, such as model requests made through an MCP client, might consume credits.
Beta and experiment features
These features are either beta or experiment and do not consume GitLab Credits.
For GitLab.com users on the Free tier, beta and experimental features don’t consume credits, but you require credits in your Monthly Commitment Pool to access them.
When a feature becomes generally available, usage of the feature starts to consume GitLab Credits on all GitLab versions and on all offerings. Beta features can change to generally available with usage billing at any time.
| Feature | Free | Premium | Ultimate |
|---|---|---|---|
| Custom flows Combine multiple agents to solve your business problems. | |||
| SAST Vulnerability Resolution Flow Automatically generate fixes and remediation steps for SAST vulnerabilities. | No | No | |
| MCP server Securely connect AI tools and applications to your GitLab instance. | No | ||
| Data Analyst Agent Analyze data and generate insights from your development metrics and project data. | No | ||
| External MCP servers Connect custom agents to external data sources and third-party services using MCP servers. | No | ||
| Knowledge Graph Create structured, queryable representations of code repositories to power AI features. | No |