GitLab Duo?

NOT REALLYA prototype of the interface is possible, but the durable product depends on difficult systems beyond the UI. Build a safer adjacent workflow instead.

BUILD FIRST

A non-sensitive or read-only version of the core experience
A manual review step wherever correctness or trust matters
A narrow companion tool that removes a real piece of work

LEAVE FOR LATER

Regulated decisions, payments, or irreversible actions
Claims of correctness, compliance, or uptime that the MVP cannot support
Sensitive integrations until security, support, and auditability are ready

RESEARCH NOTES

What this tool actually does.

Checked 2026-09-06

GitLab Duo is GitLab's AI-native feature suite embedded across the software development lifecycle, available in the GitLab UI, Web IDE, and IDE extensions for VS Code, JetBrains, and Visual Studio. It requires a Premium or Ultimate subscription plus an add-on: Core (included), Pro, or Enterprise. Features include Code Suggestions, contextual Chat, code explanation/refactor/fix, test generation, discussion summaries, code review, root-cause analysis, vulnerability explanation/resolution, and merge commit message generation. Several features are Enterprise-only; some (merge request summary, code review summary, issue description generation) remain beta. GitLab Duo has both agentic (autonomous) and non-agentic (assistive) capabilities. On self-managed instances, GitLab Duo Self-Hosted supports bringing your own LLM to keep data in-house; otherwise GitLab-managed models via the AI Gateway ar

SOURCES

RESEARCHED BUILD PROMPT

Give this to your coding agent.

Build a focused replacement prototype inspired by the GitLab Duo workflow: a dev-tool AI copilot that lives inside a Git-based DevOps platform and a desktop IDE. The core workflow: a developer edits code in VS Code-style editor; the assistant (1) autocompletes code from repo context, (2) answers natural-language questions about the current file, merge request, or pipeline, (3) explains/refactors/fixes selected code, (4) generates tests, and (5) surfaces a security vulnerability with an explanation and proposed patch, all without leaving the editor. The prototype must also support a web UI pane showing the same chat and issue context. Requirements: implement an IDE extension (VS Code) and a minimal web dashboard; a backend service with a pluggable LLM gateway (OpenAI-compatible), a code-indexing/retrieval layer (embeddings + vector store) over a local git repo; a GitLab-compatible REST endpoint for fetching MRs, issues, pipelines, and vulnerabilities (mock or real); streaming chat responses; diff preview for refactors/fixes; context window that pulls current file, open tabs, repo tree, and MR comments. Use TypeScript/Node for backend and extension; store conversation history in SQLite. Submit a working demo repo with bundled mock data. Deliberately out of scope: do not copy GitLab's brand, logo, visual design, or UI layout; do not reimplement GitLab itself (no merge request approval flows, CI execution, or container registry); no multi-tenant billing or team management; no self-hosted model orchestration. Include a README with architecture, setup steps, environment variables, and important limitations (e.g., mock GitLab data, local-only indexing, single-user assumptions).

Visit the official product page