canestackLabs

Scaffolding for AI, cloud and security.

Canestack Labs is an AI, cloud and security lab. We research and build the systems that hold up modern software: AI agents and models, the cloud they run on, and the security around both.

The Canestack mark: a cane holding up three stacked tiles The Canestack mark: a cane holding up three stacked tiles

What we work on

Three areas, one stack.

AI now runs on the same clusters, budgets and attack surface as everything else. We work across all three, because the hard problems sit where they meet.

AI

Agents, model serving and evaluation. Systems that use models safely, measure what they do, and keep the cost of every call visible.

Cloud

Kubernetes, platforms and spend. Tools that show what is running, what it costs, and what is about to break.

Security

Identity, access and secrets. Least privilege by default, sensitive values masked, and every change approved by a person.

First out of the lab

stormglass

Kubernetes, Docker and cloud cost

A native desktop client for Kubernetes clusters, Docker containers and AWS spend. Live lists, logs, terminals, YAML editing with a server dry run, cost findings with the saving for each, and an assistant that runs on the model you choose.

DockerDocker mode: the containers list with image, state, CPU, memory and ports
Cloud costCloud mode: spend, budget, forecast and waste found, with a daily spend chart (demo data)
KubernetesKubernetes mode: the Pods list of a production cluster with CPU, memory and status

Story

Why "Canestack".

Bamboo canes, lashed together, hold up skyscrapers while they are built.

Across Asia, towers still go up inside bamboo scaffolding. Each cane is light, but bound together they carry the whole building while it is built. They bend in a storm instead of breaking, and they grow back fast.

That is the job we want our work to do: hold up the systems teams build on, from the models to the cloud to the controls around them, and bend without breaking when things get rough.

The mark is the same idea. A cane holds up three tiles: AI, cloud and security.

AI

The models and agents.

Cloud

Where they run, and what it costs.

Security

Who can do what, and what is exposed.

Principles

How we build.

Your data stays yours

Our tools talk to your clusters, models and cloud accounts directly, with your credentials. There is no Canestack server in the path and no telemetry.

Research that ships

We study real problems in AI and cloud systems, then turn what we learn into tools you can run, not slide decks.

Fair prices

Priced for individual engineers, not procurement departments. Try everything first, then pay a few dollars a month or once.

Secure by default

Least privilege, secrets masked before anything sees them, and no change applied without a person approving it.

Roadmap

Where the lab is, and what is next.

  • October 2026

    Stormglass 0.2.1. A native rewrite in Rust for macOS, Linux and Windows, with cloud cost for AWS, the assistant panel, the MCP server and Docker mode.

  • Next

    More clouds. Cost for Google Cloud, Azure and DigitalOcean, in the same app and the same views.

  • Next

    Trial and Viewer. Every download starts with 14 days of everything, then a free read-only Viewer.

  • Ongoing

    AI and security research. Agents that act safely on real infrastructure, the cost of every model call, and least-privilege access by default.

Questions

What is Canestack Labs?

An AI, cloud and security lab. We research the problems teams hit where those three meet, and build tools to solve them. Stormglass is the first.

Is Stormglass your only product?

For now, yes. It is the first tool out of the lab, and the one we are putting our time into.

Where does my data go?

Nowhere we can see it. Our tools talk to your clusters, models and cloud accounts directly with your own credentials. There is no Canestack server in the path.

Do you collect telemetry?

No. No analytics, no crash upload, and no account. Licenses are signed keys checked on your machine.

Can I use local AI models?

Yes. Ollama, LM Studio, llama.cpp and vLLM are found on localhost automatically. With a local model, nothing leaves your machine.

How do I report a security issue?

Email security@canestack.com. Please include the version and steps to reproduce.

Try Stormglass on your cluster.

Every feature free for 14 days. No card, no account.