glci

Run GitLab CI/CD pipelines locally.

An experimental GitLab project.

curl -fsSL https://gitlab.com/gitlab-org/ci-cd/runner-tools/glci/-/raw/main/install.sh | bash

# Install the glci skill for your AI coding agent
glci skill install

See it in Action

Full Runner Fidelity

Uses the official gitlab/gitlab-runner Docker image. Same shell scripts, image pulls, and lifecycle as production.

Offline YAML Parsing

Resolves include:, extends:, !reference, default:, workflow: without needing a GitLab API token.

Interactive TUI

Full-screen terminal UI to browse pipeline structure, view history, and run jobs interactively.

Background Daemon

Jobs survive terminal disconnection. Detach with Ctrl+C, view logs later with glci log.

Embedded Registry

OCI registry built into the mock server. docker build/push against $CI_REGISTRY works out of the box.

Context Simulation

Test branch, merge request, tag, and environment contexts without pushing to GitLab.

Artifact Passing

Upstream job artifacts are stored and downloaded by downstream jobs via needs:/dependencies:.

Docker-in-Docker

Full DinD support with privileged mode, buildx, QEMU binfmt, and cross-platform builds.

AI Agent Skill

glci skill install teaches Claude, Cursor, Windsurf, Copilot, Duo, Codex, and OpenCode how to run your pipelines.

Remote Docker

Works with DOCKER_HOST=tcp://..., Docker contexts, and cloud-hosted Docker — no shared filesystem needed.

GitLab Pages Preview

Serve Pages artifacts locally over HTTP with glci pages serve. Preview your site before pushing.

Persistent CI Cache

Cache backed by a local Docker volume with S3-compatible API. No external bucket needed.

Quick Start

# Launch interactive TUI
glci

# Visualize your pipeline
glci show

# Run all jobs
glci run

# Run specific jobs
glci run build test

# Run a specific stage
glci run --stage test