Configure models through environment variables:
export OPENCODE_MODEL="openai/gpt-5.5"
export OPENCODE_SMALL_MODEL="$OPENCODE_MODEL"Each role can be overridden independently:
export OPENCODE_RESEARCHER_MODEL="openai/gpt-5.5"
export OPENCODE_REVIEWER_MODEL="openai/gpt-5.5"Run npm run oc-switch from a checkout, or install the published CLI globally
with npm:
npm install --global opencode-agent-orchestration-kit
oc-switchTo install the current checkout instead:
npm install --global .
oc-switchOnce a release tag exists, a checkout is optional:
npm install --global "git+https://github.com/jcarlosrodicio/opencode-agent-orchestration-kit.git#v1.0.43"
oc-switchThe TUI reads available models from OpenCode's own local opencode models --pure
catalog and discovers configured providers through OpenCode; it does not use
Pi's catalog. Startup is cache-first. Press r to request an explicit remote
refresh with opencode models --pure --refresh.
Default changes OPENCODE_MODEL, Small/title changes
OPENCODE_SMALL_MODEL, and agent rows change their corresponding
OPENCODE_<AGENT>_MODEL override. Press Enter or / to open the picker;
while it is open, all printable letters—including q, r, and s—filter the
catalog. Use Enter to apply, Esc to go back, s to save, and q to save
and exit. Space marks several agents for one assignment. The state is kept
under the XDG state directory and the model variables are updated in the
managed block of .zshrc or .bashrc.
Keep provider credentials in your shell, password manager, or OpenCode auth flow. Do not commit credentials.