Agent Pager

Agent Pager turns your Stream Deck into a control surface for your coding agents: Claude Code, GitHub Copilot (CLI and VS Code agent mode), and Cursor. When your agent asks permission to run a command or edit a file, your deck flashes and the ALLOW key shows the exact command, approve or deny without switching windows.

Included actions: • ALLOW / DENY: answer the pending permission; the key face shows the actual command, file, or URL • Open & Respond: flashes when the agent asks a question and brings it forward • Context Gauge: live context-window ring (Claude Code) • Usage Meter: session spend: Claude Code tokens, or Copilot premium requests + AI units • Toggle Speech: spoken alerts (voice, rate) • Setup Hooks: one-press install/removal of the agent hooks (auto-detected) • Voice Dictation toggle (macOS)

How it works: a small local listener plus agent hooks route permission requests to your deck. Miss the ~30 s window and the agent falls back to its own prompt, nothing hangs. Pre-allowed requests answer silently, ALLOW is always "allow once"; restart open agent sessions once after hook changes.

Away from your desk? Pair the Agent Pager iOS app (https://apps.apple.com/us/app/agent-code-pager/id6790573902) to approve from your iPhone lock screen or Apple Watch, end-to-end encrypted, opt-in, no account.

Requires Claude Code, GitHub Copilot, or Cursor (macOS 12+ / Windows 10+).

Not affiliated with Anthropic, GitHub, or Anysphere. Claude, GitHub Copilot, and Cursor are trademarks of their respective owners.

NEW: the Usage Meter now speaks Copilot. When your active session is Copilot, the tile shows premium requests and AI units consumed, Copilot's actual cost currency with a live burn sparkline, read locally from Copilot's own session records. Claude Code sessions keep the token meter; the tile follows whichever agent you're working with.

NEW: Cursor support is now first-class. This update badges its prompts and status events correctly on deck and phone, so multi-agent desks always know who's asking.

Also: the reconnect re-send from 1.3.0 has been verified live against a relay restart, a prompt parked during a brief relay blip re-delivers automatically. Reminder after updating: restart open agent sessions (all agents read their hooks at session start).