FAQ
Straight answers before you install.
A quick guide to how OpenKyrozen works, what it stores, and where its boundaries are.
What is OpenKyrozen?
OpenKyrozen is an open-source, local-first AI agent built around terminal workflows. It can work with project files, commands, Git, web research, memory, and other configured tools.
What does “self-learning” mean here?
OpenKyrozen records outcomes, corrections, tool receipts, and acceptance evidence. Repeated verified work can inform reusable skills and workflow improvements. Promotion is guarded; learning cannot grant permissions, create new runtime capabilities, approve actions, or modify model weights.
Does OpenKyrozen keep all data on my computer?
Application state is stored locally by default, with SQLite as the authoritative store. When you use a hosted model provider, the prompt content needed for that request is sent to that provider. A local model such as Ollama can be configured when you want local inference.
Which model providers can I use?
OpenKyrozen supports multiple hosted provider families and local Ollama. You can configure provider defaults and fallback behavior. See the configuration guide for the current list and setup details.
Provider configurationCan it run commands or change files without asking?
Ask and Plan modes are read-only. Agent actions go through capability checks and approval controls. Review the active mode and approval settings before granting access to high-impact tools.
Does it include a web interface?
The terminal UI is the primary experience. An optional local web UI, REST API, and MCP integration are also available.
Which operating systems and Python versions are supported?
The project documents installation for macOS, Linux, and Windows, and supports Python 3.12 and 3.13.
Is OpenKyrozen open source?
Yes. The project is released under the MIT License.
View the source on GitHubMORE DETAIL
Read the source and the docs.
OpenKyrozen’s behavior is documented alongside the code.