hivectl <command> [args]. project commands need
.hive/config.yaml (run hivectl init first); cluster-wide
commands don't. Run hivectl help for the live list.
| Command | What |
|---|---|
fire <ISSUE> --type=<T> [--repo=<URL>] [--base=<BR>] | Fire a worker against GitHub ISSUE (a number or issue-N). T: feature-implementation | bug-fix | improvement | research. Repo precedence: --repo > config's repo. |
qa --url=<U> [--target=<ISSUE>] | Fire a qa agent against a deployed URL. Optionally cross-links a target issue. |
merge <ISSUE> [--repo=<URL>] [--base=<BR>] [--force] | Merge agent/issue-<N>'s reviewed PR locally (via gh, no pod). Refuses unless the issue is status:in-review. |
status <ISSUE> <STATUS> | Set an issue's status:* label, enforcing the mutex. STATUS: ready | in-progress | in-review | changes-requested | done | needs-human. |
tail <ISSUE> | Live-tail the running pod's stdout (filtered by app=). |
logs | List this project's run dirs. |
metrics [ISSUE] | Cost / tokens / duration table. |
ui | Open the agent run-analytics UI (port auto-picks from 8001). |
expose | Port-forward prometheus + headlamp + the configured app, and start the analytics UI — one process, Ctrl-C stops all. |
| Command | What |
|---|---|
driver [--once] [--budget=<USD>] [--sleep=<s>] [--repo=<path>] | AUTO (default): unattended loop — picks the current sprint and drives it to completion, streaming output, relaunching a fresh session on each context-limit handoff / per-session budget cap. --once one session · --budget per-session $ cap (default 10) · --interactive human-driven session. |
| Command | What |
|---|---|
init | Scaffold .hive/config.yaml in cwd (idempotent). Also writes .hive/.gitignore. |
config [get <k> | set <k>=<v>] | Print resolved config; or read / edit one key. |
labels [sync | list] | Sync the canonical labels.md dictionary into the project's GitHub repo (idempotent). |
gc [--age=30m] [--all] | Delete this project's Sandbox CRs older than --age in the agents namespace. |
| Command | What |
|---|---|
local-cluster-up | Spin up the local-cluster minikube cluster. Resize at first start via env, e.g. DISK=80g MEMORY=16384 hivectl local-cluster-up. |
local-cluster-down | Delete the local-cluster minikube cluster. |
| Command | What |
|---|---|
bootstrap | One-shot fresh-machine setup: doctor → local-cluster-up → setup (stops on first failure). |
doctor | Check host CLIs and print install hints. |
setup | Install metrics-server + prometheus + sandbox CRDs + headlamp + hive-gc. |
agent-setup project | Build the pod images + create the <app>-creds k8s secret from .env. |
| Command | What |
|---|---|
version (-v / --version) | Print the hive version (from package.json). |
root | Print the resolved install dir ($HIVE_ROOT). |
help | The usage message. |