AuditKeyboard
No need for any other keyboard on your desk. This audit keyboard has one job: streamlining the auditor in the loop.
Sold out
01 / 06 · The panel
A Work Louder Creator Micro 2 Pro, flashed as an agent controller. A dial to scroll your queue, a joystick for saved views, a grid of keys for the work itself, and a wide bar you talk to. Dim white across the board means nothing is waiting for you.
02 / 06 · Add
An agent drafted something new (a finding, a control, a workflow step) and stopped for a human. Green is the safe end of the scale: approving it puts a new item into the record, and takes nothing away.
03 / 06 · Change
Something already in the record is being rewritten: a risk rating that moved, a finding whose remediation date slipped, a control description redrafted after a walkthrough. Yellow because someone already signed off on the old version, and approving this replaces it.
04 / 06 · Remove
An agent is proposing to remove an audit finding, close out a risk, or drop a control from scope. Red carries across a room because a deletion is the one change an audit trail can never quietly absorb. It sits there, lit, until a person decides.
05 / 06 · Your call
Three colors, three kinds of change, readable before you've focused your eyes. Press a slot once to arm it, again to approve. The change runs and the item lands in the register. Arm the reject key first to send one back instead. Nothing is applied until you press.
06 / 06 · Layers
Touch the strip to move between CoworkCanvas, Claude Code, Codex, and everything else. The legends change; your finger positions don't. The wide bar is the constant: it dictates into whatever has focus, on every layer.
What the keys do
Each layer turns the same keys into a different job, so the muscle memory stays put no matter which agent is running.
The frosted slot keys are the AI suggestions waiting for you. Press to approve; the change runs and the item drops into your register. Arm the reject key first to send one back. Nothing is applied until you press. Twice.
Send, approve a permission, switch modes, keep it going: drive a whole agent session from the pad. Same finger positions on both, so your hands don't care which agent is running.
Double-tap to dictate into whatever has focus; tap once to stop. The edge glows magenta while it listens. It sits on every layer, because talking to your agent shouldn't need a mode.
Straight answers
No. A Stream Deck sends shortcuts to apps. You drive it. This macropad is driven by your agents: the lights are agent state, and the keys act on a queue of pending AI suggestions. The approval flow is the product; the shortcuts are a side effect.
Claude Code, Codex, and CoworkCanvas. Layers 2 and 3 drive a Claude Code or Codex session from the pad: send, approve a permission, switch modes, interrupt a run, all using the same finger positions on both.
No. The Claude Code and Codex layers work on their own. The CoworkCanvas layer adds the approve/reject queue for teams already using that workspace.
No. CoworkCanvas is a full web workspace, and everything the pad does has a place in the interface: reviewing agent suggestions, approving or rejecting them, opening the queue, dictating. The keyboard moves the approve step off your screen and onto your desk so you can act on it without stopping what you are doing.
A Work Louder Creator Micro 2 Pro, reflashed with agent firmware. A rotary encoder, a joystick, a grid of keys with frosted caps, and a wide 2U voice bar, with per-key RGB and underglow over USB-C or Bluetooth. The keys type anywhere; the bridge that turns a keypress into an approval runs on Windows.
Sold out · next batch opens to the list first
The first batch is sold out. No charge to join. The list gets the next-batch link, the price, and an honest ship date before anyone else.
Mail client not handy? Email support@coworkcanvas.com and we'll add you.
Cowork Canvas