Run hara with no arguments for the interactive REPL, or a subcommand for one-shot work. hara --help
lists everything; hara <command> --help drills in.
Global options
These apply to most commands:
| Flag | Does |
|---|
-p, --print <prompt> | run one prompt non-interactively (headless) |
-y, --yes | auto-approve every tool action (= --approval full-auto) |
-m, --model <model> | override the model for this run |
--approval <mode> | suggest · auto-edit · full-auto |
--profile <name> | use a named profile |
-c, --continue | resume the most recent session in this directory |
--resume <id> | resume a specific session |
--sandbox <mode> | off · workspace-write · read-only (macOS) |
Core commands
| Command | Does |
|---|
hara | interactive REPL (see Interactive Mode) |
hara setup | first-run wizard: provider → endpoint → key → model |
hara init | analyze the project and write/refresh AGENTS.md |
hara doctor | check provider, auth, model, roles, skills, memory, search, MCP… |
hara -p "<task>" | run a single task headlessly |
Org & planning
| Command | Does |
|---|
hara org <task> | route a task to the role that owns it |
hara org <task> --review | loop a reviewer until the diff is clean (--rounds <n>, default 3) |
hara org <task> --review --commit | commit with an AI message after approval |
hara org <task> --role <id> | force a specific role |
hara plan [task] | decompose into atoms, sequence as a DAG, verify each |
hara plan --parallel | run independent atoms concurrently |
hara plan resume | continue a stopped plan |
See The Org Model for how roles, review chains, and plans fit together.
Code: review & commit
| Command | Does |
|---|
hara review | review uncommitted changes for bugs / security / missing tests |
hara review --staged | review only staged changes |
hara review --base <ref> | review against a base ref (e.g. main) |
hara commit | write a commit message from staged changes and commit |
hara commit -a | stage all tracked changes first |
Search & assets
| Command | Does |
|---|
hara recall [query] | search your code-asset library for snippets / playbooks (--init to scaffold) |
hara index | build the semantic index — --repo (code search), --assets, or --all (needs an embedding provider) |
Manage your org
| Command | Does |
|---|
hara roles | list roles (hara roles init scaffolds implementer / reviewer / docs) |
hara skills | list skills (hara skills init scaffolds one) — see Skills & Plugins |
hara plugin | list plugins (add / remove / enable / disable) |
hara memory | show · init · distill — see Memory & Sessions |
hara cron | scheduled tasks — see Scheduled Tasks |
hara config | set · get · path — see Configuration |
hara permissions | show / scaffold bash command rules (--init, --project) |
Sessions & integration
| Command | Does |
|---|
hara sessions | list saved sessions |
hara export [session] | render a session to Markdown (--out <file>) |
hara gateway | run a chat gateway — see Channels & Gateway |
hara mcp | run hara as an MCP server (stdio) — expose read/search tools to other clients |
hara completions <shell> | print shell completions: bash · zsh · fish |
hara login qwen | Qwen OAuth device login (free tier) |
hara enroll [url] --code <c> | join a fleet with an enrollment code |
