FlockConnectLLC

Is Ozzie open source?

Yes. Ozzie's Mac app and core are open source under the Apache License 2.0, in a public repository at github.com/tribett/ozzie. Anyone can read the code, build the app, and run it without paying. Every safety check Ozzie relies on lives in that open code, so a church running its own copy gets the same protections.

Project
Ozzie by FlockConnect, a Mac app for church staff
Published
November 14, 2026
Copyright
2026 FlockConnect
Languages
Mostly TypeScript and Rust, with some Swift
Runs on
macOS 14 or later on Apple Silicon
Supported
The latest commit on the main branch
Name and logo
Not covered by the license. The NOTICE file allows them only to describe where the work came from.

Ozzie’s own open-source page has the steps to build the app from source.

What is open, and what is hosted?

The Mac app, its shared core, the never-list and every safety check are in the public repository, and everything there works without a paid service. The hosted pieces are separate: the ozzie.church website and the managed Ozzie plans that are coming, which add a managed AI route that requires zero data retention.

In the repository
The Mac app, the shared logic and its tests, the native checks that run on every request before it leaves the Mac, Planning Center sign-in and the fixed read requests, and the never-list.
Hosted, not in it
The ozzie.church website, and the managed plans: a managed AI route with zero data retention, no keys to manage, and help getting started.
Without the service
Use your own AI provider, a model on your own Mac, or no AI at all. Every check and draft still works.

Is FlockConnect open source?

No. FlockConnect, the Church Relationship Manager, is closed source and runs as a hosted service at flockconnect.com. Ozzie is the company’s open-source project. The two are separate products, and Ozzie is a standalone Mac app credited as Ozzie by FlockConnect to show who makes it.

How can I contribute to Ozzie?

Open an issue on GitHub before starting anything larger than a small fix, so the approach can be agreed first. Use fixed, fictional data in tests, issues and screenshots, never a real person's details. Contributions are licensed under Apache-2.0, and a change that weakens one of Ozzie's safety rules will not be merged.

The rules that are not up for change:

  • Ozzie only reads Planning Center. There is no write path and no tool that sends a message.
  • The names Ozzie reads never go to an AI model. Background-check facts never go to a model at all, even as placeholders.
  • No scoring or ranking of people, and no labels such as "at risk".
  • No generated prayers, sermons, devotionals or scripture, and nothing that speaks to a church member as if it were a pastor or a friend.
  • No secrets in the repository.
  • Every safety check lives in the open repository.

Before a pull request, run the checks CI runs: the type check, lint, the TypeScript tests and the native tests. A change to what Ozzie reads from Planning Center, to what is sent to a model, or to any rule above needs a second, independent review. The full guide is CONTRIBUTING.md, and the place to start is the project’s issues.

How do I report a security problem in Ozzie?

Privately, never in a public issue. Use GitHub’s private reporting on the repository, or email [email protected]. Include what you found, how to reproduce it, and which version or commit you used, with fictional data in place of a real person's details. Ozzie is maintained by a very small team, so there is no bug bounty.

Reports in these areas are especially welcome:

  • any way Ozzie could change or send something in Planning Center or another church system;
  • any way a name Ozzie read, a contact detail or a background-check fact could reach an AI model or another app;
  • any way text inside church data, such as an email, a note or a plan title, could be followed as an instruction;
  • sign-in, token storage in the macOS Keychain, and the connection between the Mac app and ozzie.church.

The project’s security policy is SECURITY.md. For FlockConnect itself the same address applies, and the reply times are in its vulnerability disclosure policy.