Command line · 07
Connect cloud KataGo to GUIs and scripts with Read19 CLI
Sign in once with the native CLI, then choose GTP or Analysis Engine JSONL. Local clients do not manage a model or GPU.
Install the native CLI and sign in
Download the native macOS, Linux, or Windows build from /cli and run read19 login; use read19 login --no-open on a headless machine. Then run read19 engine setup --name "My Go GUI" to create and store a dedicated access key.
Login authorizes the account; engine setup creates a separately revocable credential for this client. Select engine gtp in a GUI
In Sabaki, q5Go, or Lizzie engine settings, point the executable to read19 and set arguments to engine gtp. Probe compatibility with list_commands or known_command. Read19 implements a practical KataGo GTP subset, not every extension.
The GUI sees a local GTP process. The CLI translates positions into cloud requests and returns GTP responses. Select engine analysis for scripts
For complete machine-readable JSON, run read19 engine analysis, write one Analysis Engine query per line to stdin, and read results line by line from stdout. Use read19 usage, statement, and status to inspect usage, ledger entries, and service health.
The JSONL path bypasses a GUI and fits shell pipelines, Python, Node.js, and batch research tools.
Full command reference
Signs of success
- read19 account can load the account
- GUI protocol_version returns 2
- Candidate updates keep arriving
- Usage shows the matching job