Home / Case Studies / Building With Brajko
Case Study

We Built This Website With Our Own OpenClaw Agent

Our OpenClaw agent Brajko creates pages, deploys tracking scripts, and pushes code to production - all from a Slack message. This is the story of how we built the site you're reading right now.

Slack
Single Interface
Auto
Deployment
0
Manual Deploys
24/7
Available
Publish Flow
Slack request -> Brajko edits code -> GitHub push -> Firebase auto deploy -> live page
Video

Watch the build story

The Challenge

We needed a website. We also needed proof our own product actually works.

When we decided to launch GO AI Agent as an OpenClaw services brand, the first problem was obvious: we needed a website that clearly explains the capabilities, services, and delivery model.

The better opportunity was louder. If we sell autonomous AI assistants to other businesses, the strongest proof is to use one ourselves in real production work, not in a sandboxed demo.

So we built the website with our own OpenClaw agent, Brajko, and then handed ongoing page requests to it through Slack. The goal was simple: anyone on the team should be able to ask for a new page, a script deployment, or a content change without opening a hosting dashboard.

Outcome Target

  • Requests handled from Slack
  • No codebase hunting for routine changes
  • Production deploys within minutes
  • One system proving the product itself
The Build

From Codex to Brajko: how the site came together

1. Scaffold the site

The initial site was scaffolded with Gemini for visual direction and Codex for development. We focused on SEO structure, semantic markup, page templates, and a responsive layout that stayed simple enough for an agent to navigate.

The result is a clean vanilla HTML, CSS, and JavaScript site. No heavy framework, no build complexity getting in the way of real edits.

2. Connect the pipeline

We pushed the site to GitHub and connected the repo to Firebase App Hosting. Every push to the main branch goes live within minutes.

That made the delivery path boring in the best possible way: small code change, GitHub push, automatic production deployment.

3. Give Brajko access

Using a scoped fine-grained GitHub token, we gave Brajko access to the website repository and nothing else. It can read the codebase, create and edit files, commit changes, and push to the main branch.

That matters because the agent is not just drafting suggestions. It is changing the production system, and the GitHub-to-Firebase pipeline makes those changes visible on the website within minutes.

If you want the broader service model, start with services, then look at the rest of the case studies, and if you want to see the agent page tied to this build, open Brajko. For a lighter example of the same design system, open the demo page.

Deployment Pipeline
Slack
Team sends request to Brajko
Brajko
Writes code and commits changes
GitHub
Pushes to main branch
Firebase
Auto deploys to production
Live
Visible on goaigent.io
Capabilities

What Brajko does on this website

Page Creation

Builds new pages from scratch, following the existing design system and content structure. One Slack message, full page delivered.

GTM Deployment

Deploys Google Tag Manager scripts, creates tags, triggers, and variables directly - no need to open the GTM interface.

Code & Deploy

Edits HTML, CSS, and JS files, commits to GitHub, and triggers automatic production deployment through Firebase.

Third-Party Scripts

Installs and configures tracking pixels, analytics tools, consent managers, and other third-party scripts on the website.

Sitemap Management

Generates and updates the XML sitemap as new pages are added, keeping search engine indexing current.

Context-Aware

Knows the team, the brand, and the project goals. Outputs match our design language without needing a detailed brief every time.

"He knows our team. He knows what we want to achieve. He's deeply briefed about all our projects. It's essentially like talking to a teammate - not an AI agent that doesn't know anything."

Stefan, AI Workflow Builder at GO AI Agent

The Experience

What it actually feels like to work this way

Task Before Brajko With Brajko
New page 1-2 hours with a developer One Slack message and a few minutes
Deploy GTM script Open GTM, edit code, deploy Slack command and auto-deploy
Content update Open file, edit, commit, deploy Describe change in Slack and it goes live
Who can make changes Developers only Anyone on the team
Tech Stack

What powers this setup

OpenClaw Firebase App Hosting GitHub Slack Google Tag Manager Codex Gemini

Why we kept it simple

The site is vanilla HTML, CSS, and JavaScript. No React, no build system, no CMS. That is deliberate. A simple codebase means the OpenClaw agent can read, edit, and deploy without wrestling framework complexity.

The infrastructure is equally straightforward: GitHub for version control, Firebase for hosting, and Slack as the command interface.

Security scope

Brajko's GitHub token is a fine-grained personal access token with access limited to a single repository. It cannot access other projects, organization settings, or sensitive configurations.

Every commit is traceable to the agent in the Git history, which is the part most teams pretend they will do later and then never do.

Why This Matters

The site you're reading is the proof

Most agencies selling AI services show you a demo environment. We show you the actual product, running in production, on the website where we present our services.

The pages Brajko creates are the pages you browse. The scripts Brajko deploys are tracking your visit right now. The code Brajko pushes is the code serving this page.

This is what we mean when we say we build AI assistants that actually do things. Not a concept. Not a slide deck. A working agent, doing real work, in production.

And if we can build an agent that manages its own website - creating pages, deploying analytics, pushing code - imagine what it could do for your calendar, your project management, your lead pipeline, or your reporting.

FAQ

OpenClaw case study questions

Can an OpenClaw agent build a website?

Yes. Brajko can create new pages, follow the existing design system, commit the changes, and push them live through the repository and hosting pipeline.

How does an OpenClaw agent deploy code?

A Slack request becomes a code change in the repo, the change is pushed to GitHub, and Firebase App Hosting deploys it automatically.

Is Brajko just suggesting edits?

No. Brajko has direct access to the codebase and can make real production changes instead of only drafting recommendations.

Want an AI assistant that actually does things?

Book a 30-minute demo. We'll show you what Brajko does for us - and map out what yours could look like.