FAQ
Is Kanna free?
Section titled “Is Kanna free?”The Kanna software itself is free and open source. Underlying provider costs (Claude, Codex, OpenRouter) depend on your account.
Does Kanna upload my code anywhere?
Section titled “Does Kanna upload my code anywhere?”No. The agent runs locally — claude or codex CLI subprocesses on your
machine. Only the prompts and tool outputs you explicitly send go to the model.
Telemetry (turn timings, memory, run counts — never prompt or code content) is the one thing that leaves by default, and it is a single switch under Settings → General → Telemetry Tracing.
Which Claude plan does Kanna bill against?
Section titled “Which Claude plan does Kanna bill against?”Whichever plan the OAuth token you registered belongs to. Kanna passes your token through to the Claude Agent SDK; add one in Settings → Providers.
Can I use both Claude and Codex in the same chat?
Section titled “Can I use both Claude and Codex in the same chat?”Yes — switch providers mid-chat from the composer. Previous turns remain unchanged; the new turn uses the picked provider.
Where is my data stored?
Section titled “Where is my data stored?”~/.kanna — chats, transcripts, settings, and your OAuth tokens. The path is
not configurable; it follows HOME, so relocating it means running Kanna with a
different HOME. See Self-host basics.
How do I set the port or a password?
Section titled “How do I set the port or a password?”They are CLI flags, not env vars: kanna --port 8080 --password <secret>.
KANNA_PORT and KANNA_PASSWORD do not exist and never have.
Can I run Kanna headless?
Section titled “Can I run Kanna headless?”The server runs headless, but Kanna is a web UI — a browser is required to interact with it. For automation, drive the Claude/Codex CLIs directly, or use cron jobs and loops inside Kanna.
Windows support?
Section titled “Windows support?”Not supported. macOS and Linux only; WSL works in practice but is not tested.
Can several agents work at once without colliding?
Section titled “Can several agents work at once without colliding?”Yes, if each has its own checkout. Starting work from a board card creates a branch and a git worktree per card, which is what makes three concurrent agents safe. Several chats in the same directory will happily overwrite each other’s edits.