CanaStack
Sign in
AI-Native Architecture Platform

CanaStack

Describe your project. AI builds the architecture.

Sign in
⬡
Cubes
Visual architecture
⚡
Multi AI
Generated code
🚀
Deploy
Isolated containers

Describe your app. CanaStack builds, deploys and maintains it.

An AI-powered platform that turns an idea into real microservices, live on your own subdomain, with self-healing and usage-based billing.

💬
Describe
Explain your project in plain language. The AI proposes a "cube" architecture.
⬡
Visualize
Tune the architecture via graph, chat or analysis — all in sync.
🚀
Deploy
Each cube is coded, built and deployed in its isolated container.
🌐
Go live
Reachable on your subdomain or your own domain, auto HTTPS.

The platform in pictures

🔒console — Architecture
Cubes
⬡Frontend
⬡API
⬡Database
⬡Auth
Architecture graph

Everything you need to ship

From a plain-language idea to a live, maintained product — the whole stack is handled for you.

AI code generation
Describe an app in chat — the AI writes complete, working microservices.
Auto-heal
Build and runtime errors are caught and the AI rewrites the code until it runs.
Isolated containers
Every project runs in its own secure, resource-limited container.
Instant public HTTPS
Each project goes live on its own URL with an automatic SSL certificate.
Dev, staging & prod
Isolated environments with one-click fast-forward promotion to production.
Project editor & logs
Inspect generated files, prompts, AI logs, tasks and deploy output from the dashboard.
Per-project Git
Every project and service gets its own Git repository, versioned automatically.
SSO & auth built-in
Keycloak single sign-on, 2FA and user management wired in from day one.
Pay per use
Pay for AI by the call and servers by the minute — top up with PayPal or gift cards.
AI agents
A team of AI agents builds and maintains your project: a master agent holds the architecture and delegates to specialised code, git, infrastructure and chat agents. See them, wire them and edit their prompts on a live graph.
Agents as endpoints
Turn any agent into an OpenAI-compatible endpoint your own application calls. Usage is billed to the project, with cost, call count and recent exchanges per agent.
Custom prompts and models
Write each agent’s prompt yourself and pin the exact model it runs on. An unavailable model fails plainly — your agent is never silently answered by a different one.
Open by design

Not a closed black box

CanaStack builds real software you can inspect, change, export and run elsewhere. The AI accelerates the work; you keep access to the code, repositories, Docker files, configuration and deployment path.

Source code access

Every project is backed by Git repositories. Review commits, inspect generated files and keep the history.

Docker & config

Services are deployed from real container definitions and environment configuration, not hidden builder state.

Bring your runtime

Run on managed CanaStack containers, your own VPS, a colocated box or a machine at home over SSH.

Logs, files & memory

Project files, AI logs, task history, memory and deployment output stay visible inside the platform.

Deployment options

Managed when you want it, portable when you need it

1
Managed containers

Launch isolated CanaStack production containers with automatic HTTPS and per-minute compute billing.

2
Your VPS or bare metal

Register a remote host, store SSH credentials encrypted and deploy projects there.

3
Home machine

Use your own always-on computer or lab server when that is the right place to run.

4
Domains and routing

Use a free subdomain, connect your domain or register one with private WHOIS.

Chat Agent, multi-agents, A2A and MCP

The platform is built around agents that understand the project, the runtime and the operational tools around it.

Chat Agent

One conversation that can act

Ask for features, fixes, deploys or explanations. The agent reads context, edits code, queues builds and verifies results.

Multi-agent

Specialized project agents

Create agents for product, backend, frontend, operations or review work. Keep them private by default.

A2A

Projects can talk to projects

Link trusted projects so one project agent can ask another for data, signals or coordinated changes.

MCP

Platform context for AI

Agents receive structured context about architecture, files, capabilities and platform services instead of guessing.

Review

AI review before deploy

Generated changes are checked against contracts, runtime wiring and build output before they reach production.

Routing

Provider routing and fallbacks

Use the right engine for analysis, code, review or chat, with quota-aware fallbacks and audit logs.

Ready to build?

Create your account and ship your first project in minutes.

Get started free