Glaboud

Close your laptop—Glaboud's Gitlab powered AI engineering team keeps shipping code from the cloud under your supervision.

Comment

Glaboud — Your AI Engineering Team, Always Running, Always in Your Pocket

Mobile-first. Context-first. GitLab-native.

Manage autonomous Designer, Implementor, Reviewer, and Debugger agents from anywhere while Glaboud's repository intelligence layer minimizes context drift, reduces token consumption, and keeps software shipping even when your laptop is closed.

Inspiration

Software development is still fundamentally tied to a laptop.

Developers spend hours managing repositories, reviewing code, fixing bugs, maintaining documentation, and repeatedly providing context to AI coding tools. Whether you're commuting, attending a conference, traveling, or simply away from your desk, productive development often slows down because the development environment remains locked to a workstation.

At the same time, modern AI coding assistants face another challenge: context drift. As repositories grow, agents repeatedly consume large portions of the codebase for every task, increasing token costs while gradually losing awareness of architectural decisions, coding standards, and project conventions.

We believed there had to be a better way.

What if software development could continue even when your laptop was closed?

What if developers could supervise an entire engineering team from their phone while autonomous agents worked in the cloud?

That idea became Glaboud.

What It Does

Glaboud is a mobile-first, cloud-native software engineering platform that transforms GitLab repositories into autonomous engineering workspaces powered by a collaborative swarm of specialized AI agents.

Instead of relying on a single coding assistant, Glaboud deploys four dedicated engineering agents:

Designer

Plans system architecture, analyzes requirements, evaluates technical trade-offs, and creates implementation strategies.

Implementor

Builds features, writes code, executes development tasks, and converts specifications into working software.

Reviewer

Performs automated code reviews, validates correctness, enforces engineering standards, and improves maintainability.

Debugger

Investigates failures, traces root causes, reproduces issues, and resolves bugs before they reach production.

Each agent runs independently inside its own isolated containerized environment, enabling parallel execution while maintaining reproducibility and security.

Through deep GitLab integration, agents can interact directly with repositories, issues, merge requests, branches, documentation, and CI/CD pipelines, allowing them to participate in real-world software engineering workflows.

Most importantly, developers can supervise the entire process from a mobile device. Launch tasks, monitor progress, review outputs, approve merge requests, inspect documentation, and guide development from anywhere while your AI engineering team continues working in the cloud.

The Problem We Solve

Modern AI development tools suffer from three major limitations:

Laptop Dependency

Development slows down whenever developers leave their workstation.

Context Drift

AI agents repeatedly ingest large portions of repositories, gradually losing awareness of architectural decisions and project conventions.

Token Inefficiency

Every new task often requires processing large amounts of repository data again, increasing latency and operational cost.

Glaboud addresses all three.

Our Innovation: Context-First Engineering

Most AI coding platforms treat repositories as raw code that must be repeatedly loaded into context windows.

Glaboud takes a fundamentally different approach.

When connected to a GitLab repository, Glaboud performs deep repository analysis and automatically generates structured, navigable project documentation.

This documentation becomes a living knowledge layer that captures:

  • System architecture
  • Module relationships
  • APIs and interfaces
  • Engineering conventions
  • Repository structure
  • Dependency mappings
  • Domain-specific knowledge

Instead of loading thousands of files into every prompt, agents retrieve only the documentation and code sections relevant to the task they are performing.

This targeted retrieval architecture delivers:

  • Lower token consumption
  • Faster execution
  • Reduced operational costs
  • Better architectural consistency
  • Less context drift
  • Improved scalability for large repositories

The documentation continuously evolves alongside the codebase and remains accessible to both developers and agents through the Glaboud interface.

The result is an engineering system where agents spend less time searching for context and more time shipping software.

Mobile-First Software Development

Traditional development assumes developers are sitting in front of a laptop.

Glaboud removes that assumption.

Our mobile-friendly web application allows developers to:

  • Create and assign engineering tasks
  • Monitor agent activity in real time
  • Review implementation progress
  • Approve or reject merge requests
  • Track CI/CD pipeline status
  • Navigate repository documentation
  • Intervene whenever human judgment is required

Whether you're attending a conference, commuting, traveling, in a meeting, or simply away from your desk, development continues uninterrupted.

Instead of carrying a laptop everywhere to keep projects moving, developers can manage an entire engineering team from their pocket.

Powered by GitLab

Glaboud is built around the rich DevOps ecosystem of GitLab.

Agents can interact directly with:

  • Repositories
  • Issues
  • Merge Requests
  • Branches
  • Wikis
  • Project Documentation
  • CI/CD Pipelines

Every generated change flows through established GitLab workflows, ensuring that code quality, testing, review, and deployment remain governed by existing engineering controls.

Rather than bypassing software engineering best practices, Glaboud amplifies them through autonomous execution.

This ensures that no code reaches production without passing through the same quality gates trusted by professional engineering teams.

How We Built It

Core Components

  • Google Cloud Agent Builder (Agent Studio + ADK)
  • GitLab-MCP Integration
  • Autonomous Multi-Agent Orchestration Engine
  • Containerized Agent Runtime
  • Repository Intelligence Engine
  • Documentation Generation System
  • Mobile-First Web Application

Agent Infrastructure

Each agent operates inside an isolated Docker container, enabling secure and parallel execution across engineering tasks. Agent workflows were designed and refined using the Google Cloud Agent Builder ecosystem, combining rapid iteration in Agent Studio with custom orchestration through ADK.

Repository Intelligence Layer

Repositories are ingested and transformed into structured documentation that can be navigated by both humans and AI agents.

This shared knowledge layer dramatically reduces unnecessary context loading, minimizes context drift, and improves task accuracy.

Engineering Workflow Integration

Agents operate directly within GitLab workflows, allowing autonomous development without sacrificing review processes, testing standards, or deployment safeguards. Built on Google Cloud Agent Builder, Glaboud combines AI-native workflows with GitLab's proven software delivery ecosystem.

Challenges We Ran Into

Building AI agents that write code is relatively straightforward.

Building AI agents that remain effective across large, evolving repositories is significantly harder.

Some of our biggest challenges included:

  • Preventing context drift across long-running tasks
  • Reducing token usage while maintaining accuracy
  • Coordinating multiple specialized agents
  • Generating useful repository documentation automatically
  • Maintaining consistency across engineering workflows
  • Designing a practical mobile-first development experience

These challenges ultimately led to Glaboud's context-first architecture, which became one of the platform's strongest differentiators.

Accomplishments We're Proud Of

  • Built a specialized multi-agent engineering system instead of a single coding assistant
  • Created a context-first repository intelligence layer
  • Enabled mobile-first supervision of autonomous software development
  • Integrated agents directly into GitLab workflows
  • Reduced repository context overhead through targeted retrieval
  • Designed a scalable architecture for large codebases

What We Learned

The biggest lesson was that better software engineering AI is not solely about larger models.

Context management, specialization, workflow integration, and engineering discipline often matter more than raw model capability.

We also learned that developers do not necessarily want AI to replace engineering teams. They want AI systems that operate within existing workflows, respect engineering best practices, and help teams move faster without compromising quality.

What's Next for Glaboud

Our vision is to evolve Glaboud into a fully autonomous engineering cloud.

Future plans include:

  • Long-term repository memory
  • Cross-project organizational knowledge
  • Technical debt management agents
  • Release planning and coordination agents
  • Predictive debugging and incident prevention
  • Team collaboration workflows
  • Enterprise-scale GitLab deployments

Ultimately, we envision a future where software projects continue progressing around the clock, powered by specialized AI engineering teams that deeply understand the codebases they work on.

Glaboud transforms a GitLab repository into an always-on engineering organization—one that keeps building, reviewing, documenting, and improving software even when the developer's laptop is closed.

Built With

Share this project:

Updates