Open source · MIT licensed · Built in public

Software and research that anyone can use, inspect, and run themselves.

Built for Roofers and anyone with numbers to run.

Maris Stella Labs is an open portfolio of business software and research: focused tools that run entirely in your browser, self-hostable apps a small business can deploy into its own cloud account, and reproducible research with its code and tests. No accounts, no paywalls, and no data leaving your hands.

Browser tools
1062
Self-hostable apps
67
Command-line tools
192
Research projects
1059

How it works

Free tools you can use, check and own

  1. 1

    Open a tool and type your numbers

    1062 tools run entirely in your browser. No account, no upload: the arithmetic happens on your device, and works offline once opened.

  2. 2

    Check how the answer is worked out

    Every formula, test and research result is open source under the MIT licence, so you can read exactly how a number was reached.

  3. 3

    Run your own copy

    67 self-hostable apps deploy into your own cloud account with your own keys. Nothing of ours sits in between. Deploy your own →

Who it is for

Find the tools for your trade or business

The same tools, filed by who they are for. Each page lists the calculators that kind of business uses, and what to price first if you are just starting one.

Browse

Tools by the job they do

1321 free tools in 15 sections. Each section's page lists every tool in it, grouped and filterable. See every section →

By type

Apps you run yourself, and checks for your terminal

See it working

Real answers, not demos

Every picture here is a tool on this site with its example loaded. Open one and it does the same for your numbers, in your browser.

Recorded from the live tools: each one loads its own worked example and shows the answer.

Open research

1059 reproducible research projects

Verification markets, scoring rules, decentralised training, chip design, energy and robotics. Each project states its question, ships runnable code and tests, and writes up what it found and where it falls short. Browse all research →

Proper scoring & elicitation 28 Effort & peer prediction 9 Verification games & slashing 19 Detection & tolerances 14 Markets as routers 13 Prediction-market design 9 Paying contributors 8 Decentralised training systems 23 Digital twins & sim-to-real 164 LLM chip design loops 40 Robot manufacturing loops 53 Ambient AI assistant 46 Self-replicating manufacturing 34 Simulation-first design 64 Engineered clothing 45 Task-specific AI accelerators 40 Algorithmic chip-design loops 42 Open-source purpose-built robots 44 The software-engineering loop 43 Adversarial human imitation 41 Open-source residential construction 41 Autonomous multi-angle capture 34 Sensing, imaging and 3D 31 Energy supply and dense energy storage 39 Robot societies among people 31 Agentic co-design of low-cost robots 27 Laundry-hanging robots 25 Machine learning on qualitative data 23 Robots on the building site 27 Civilizations & group psychology 1 History & human sciences 1

Writing

Notes from building it in public

Essays drawn from this repository's own history and numbers. Read all 1597 posts →

Also here

Run your own copy, read the guides, follow the work

Deploy your own copy

Every app and tool is MIT licensed. Pick one, paste your own cloud credentials, and get a running copy in your own account. Nothing of ours sits in between.

Deploy your own →

Guides and comparisons

The explainers written beside each tool, and products compared side by side, filed by trade.

Read the guides →

Built in public

What shipped this week, what is in progress, where the project is going and how AI agents build it, from the repository's own records.

Updates and roadmap →