magpie

Agents

Profiles
Appearance
Follows the Omarchy theme, as picked in Omarchy's menu
Colorblind mode
Language
Text size
Click the tray icon to
Provider in model names
Codex subagents on any model
A GPT lead spawns subagents on magpie’s other models. New threads only; Codex’s own multi_agent_v2 still wins
Long conversations
Codex and Claude Code compact at 272K (non-Claude [1m] too) or the size typed; full window: slower, costlier
Codex thread titles
Codex thread descriptions
Codex composer predictions
Codex auto-review model
Open at login
Prevent sleep
Keeps this computer awake while magpie runs tasks
Allowances in the menu bar
Menu bar cost
What magpie's requests cost, beside their tokens in the menu bar and its tooltip; off, the tokens alone
Allowance display
How much of each window is used, or how much is left, on the Usage page, the panel and the menu bar
Allowance reads
When I ask: opening or refreshing the Usage page, a card's refresh, and magpie quota or accounts read them; routing, the menu bar and alerts go by the last reading till then. An account a vendor turns away for its quota is still read.
Currency
Allowance alert
Low balance alert
Reset reminder
Task done alert
Warm up on reset
Daily warm-up
Warm up on reset
Daily warm-up
Daily check-in
Daily check-in
Daily check-in
Daily check-in
Proxy
GitHub
Gateway
Model ids without the provider
The gateway’s model list also names each model by its own id, as gpt-5.6-sol beside codex/gpt-5.6-sol, for apps moved from another proxy. One that several providers serve goes to them in the Providers order
Local network
Web pages
Images
Web search
Replies
Disabled models
Agents' sign-ins