Commands · the /mooter namespace
Every command is a real Claude Code slash command.
Mooter registers under the /mooter namespace — typed right in your Claude Code session, tab-completed, never clashing with the built-ins. No new CLI to learn. Click any command below to copy it.
claude · in-session
> /mooter why
last prompt → T2 sonnet
├─ intent arch · complexity med
├─ pack diagram-systems (trust 98)
└─ over opus → saved this call
claude · in-session
> /mooter override T3
next prompt pinned → T3 opus
└─ routing paused for 1 turn · you're the boss
core
Hook into Claude Code — writes the config and verifies the connection.
Explain the routing call for the last prompt: tier, model, and the reason.
Print the live statusline — savings, tier, budgets, adapter.
Rate the last answer. Feeds the routing quality signal.
<tier|model>
Force the next prompt onto a specific tier or model.
Your savings digest — what routing saved you, by pack and tier.
packs
List the packs installed on this machine.
<id>
Show a pack: skills, MCPs, agents, trust score.
<query>
Search the registry by domain or keyword.
<id>
Install a pack and map it into routing.
Publish your own pack to the registry.
forge · local adaptersWave 5 · in development
Adapter training state — idle, training, or active.
Train a local DoRA r=32 adapter on your codebase. ToS-safe, overnight.
Evaluate a trained adapter against the blind judge set.
adapterWave 5 · in development
<id>
Activate a trained adapter for routing.
Deactivate the current adapter.
privacy
Toggle opt-in telemetry. Default OFF. Prompts are hashed, never sent in plaintext.
Sub-commands shown flat for reference; in-session, /mooter tab-completes the whole tree. Forge and adapter commands ship with Wave 5 (in development) — the rest are live today.
Community project · not affiliated with Anthropic.