Keyboard shortcuts

ShortcutAction
Alt+P

Summon or dismiss Pluma from anywhere

Enter

Submit query or command

Tab

Toggle the chat panel (Whisper) open/closed

Escape

Dismiss window (keeps agent running if active)

Shift+Enter

New line in chat input

Interface

Pluma has three layers:

LayerDescription
Command bar

The main input — always visible when open. Type queries or slash commands here.

Whisper panel

Full chat history. Opens with Tab or when a chat response arrives. Has its own send field for follow-ups.

Settings

Click the ⚙ icon to set your OpenRouter API key and model.

Slash commands

Type a slash command in the command bar. All commands act on clipboard content unless you provide a direct argument.

CommandWhat it does
/polite

Rewrites clipboard text as professional and warm

/concise

Removes fluff, shortens while preserving meaning

/tldr

Summarises to exactly 3 bullet points

/bash

Translates description to a PowerShell command

/regex

Converts plain English to a regex pattern

/explain

Explains code or text in simple terms

/doc

Generates docstrings/comments for code

/translate [lang]

Translates to the specified language (default: English)

Copy/Chat mode: The AI decides whether to paste the result directly to your cursor ([COPY]) or open a chat panel ([CHAT]). Transforms paste; questions chat.

Clipboard context

Every time Pluma opens, it reads your clipboard. That content is shown as a "ctx" strip in the chat panel and automatically included in all requests as context. You don't need to paste or mention it.

Triggering agent mode

Agent mode activates automatically when your query contains action keywords:

open · click · search · find file · run · go to · browse · navigate read file · list files · my desktop · my documents · install · download move · copy · delete · rename · create file · write to · automate

You can also trigger it explicitly from the Whisper chat panel — same keywords apply.

Agent tools

ToolDescription
take_screenshot

Captures the screen. Agent calls this before clicking anything.

mouse_click(x,y)

Clicks at absolute pixel coordinates. Shows targeting reticle.

double_click(x,y)

Double-click at coordinates.

type_text(text)

Types at the current cursor position.

key_press(key)

Presses a key or combo: enter, ctrl+c, ctrl+v, alt+tab, win, f5…

scroll(x,y,amount)

Scrolls at coordinates. Positive = down.

run_terminal(cmd)

Runs a PowerShell command. Returns stdout, stderr, exit code.

read_file(path)

Returns full text content of a file.

write_file(path, content)

Creates or overwrites a file.

list_directory(path)

Lists files and folders at the given path.

recall(query)

Searches conversation memory for relevant past context.

Computer use

When performing UI tasks, the agent always takes a screenshot first, identifies pixel coordinates of the target element, then acts. A cyan targeting reticle appears at every click location.

For web tasks, Pluma opens Chrome using start chrome "url" via PowerShell. Make sure Chrome is installed.

API key

Pluma uses OpenRouter for all AI calls. Get a free key at openrouter.ai/keys — many models are free-tier. Your key is stored in the OS keychain (Windows Credential Manager), never in plain text.

Memory & sessions

All conversations are saved to a local SQLite database in your AppData folder. The session resets automatically at midnight each day. The agent can search this memory using the recall tool.

To clear manually: press Escape to dismiss, then delete %APPDATA%\com.pluma.app\pluma.db.