Built in the open.
Wizard is a local-first, autonomous data analysis agent — a manager model that plans and a worker model that writes code, investigating a real dataset in a sandboxed process until it has an answer it can independently verify. It's built as an open project: the code, the reasoning behind it, and the way it's governed are all public.
Open source
Wizard is released under the BSD 3-Clause license — a permissive license. You can use it, modify it, and ship it in a commercial product without asking permission or paying anything; the only real condition is keeping the copyright notice and disclaimer intact, and not using the project's name to endorse a derivative without asking. Like almost all open-source licenses, it comes with no warranty.
Read the full license on GitHubContributing
Contributions to the backend, frontend, or CLI go through the same review path as everything else: fork, branch, and a pull request against master. Commits follow the Conventional Commits format, and it's enforced automatically — CI won't pass without it. Changes to anything that executes generated code (the sandbox, the code guard, session handling) are held to a higher bar and need tests that demonstrate the new boundary actually holds, not just that the happy path works.
Read the contributing guide on GitHubSecurity
Generated code runs in an OS-contained subprocess or a locked-down Docker container by default — that boundary is treated as the project's most sensitive surface. If you find a way through it, or any other vulnerability, the ask is simple: don't open a public issue or post a proof of concept. Report it privately first, so there's a fix out before the details are.
Read the full security policy on GitHubCommunity standards
The short version: be someone people want to collaborate with. Disagreements about code and design are normal and welcome; personal attacks, harassment, and bad-faith arguments aren't. The full policy is adapted from the Contributor Covenant, and it's the same standard for everyone — maintainers included.
Read the code of conduct on GitHub