One agent for every cloud

Every answer cited. Every architecture exportable. Runs on your machine — and already knows AWS, Azure and Google Cloud.

$docker compose up -d Get started →
AWS Available
Azure Available
Google Cloud Available
live demo · drag to orbit · scroll to zoom · hover a service
🔒 Coming soon
All three clouds are live. Pick a provider and explore a real 3D architecture, grounded in that provider’s docs.
↑ a real AgentCloud 3D architecture (serverless web app on AWS). This exact view is generated for any question in the app.

What AgentCloud does

Not a chatbot that guesses — a grounded agent that cites its sources and visualizes the build.

📑

Ask → cited steps

Every answer is built from your cloud’s official docs & whitepapers, and every step links to the exact source page. No hallucinated APIs.

🧊

Design → live 3D

Say “design a data lake” and get components, data flow, trade-offs, clarifying questions — plus an interactive 3D architecture you can orbit.

🔒

Private & local

Runs on your machine with an embedded vector store. No Docker, no database to set up. Your questions stay yours.

💬

Terminal or web

Work in your shell with agtcld, or open the immersive web app. Answer the agent’s follow-ups to refine a design in place.

🧩

Real topology

Diagrams reflect the actual architecture — fan-out, tiers, monitoring and scaling links — with real service icons, not a straight line.

☁️

Multi-cloud

One workflow across AWS, Azure and Google Cloud — all three live today, same agent, same 3D designs.

How it works

Index

Point it at the cloud’s docs (or your own). It chunks + embeds them locally into an embedded vector store — GPU-accelerated.

Ask or design

Ask a how-to, or “design …”. It retrieves the right sources across services and composes a grounded, cited answer.

Visualize

Designs render as a live 3D architecture — real service icons, animated data flow, labeled connections you can orbit & zoom.

Get started

Two commands and it’s running — nothing to install, nothing to configure. It arrives already knowing AWS, Azure and Google Cloud.

# 1. grab the compose file
curl -LfO https://raw.githubusercontent.com/agentcloudai/agentcloud/main/docker-compose.yml

# 2. start it
docker compose up -d

# then open http://localhost:8000 and paste your OpenAI key when it asks

The first launch takes a few minutes while it downloads itself and everything it knows about AWS, Azure and Google Cloud. After that it opens instantly — and runs entirely on your machine, so your questions and your key never leave it.

Prefer Python? pip install agtcld