Real-time agent tracking dashboard for the GitHub Copilot CLI
Homebrew (recommended on macOS):
brew tap DUBSOpenHub/copilot-cli-agent-pulse
brew install agent-pulseQuick installer (any platform):
curl -fsSL https://raw.githubusercontent.com/DUBSOpenHub/copilot-cli-agent-pulse/main/quickstart.sh | bashThen just type agentpulse — the dashboard opens in a new window of your current
terminal emulator (auto-detects Ghostty, iTerm, Kitty, WezTerm, Alacritty, Warp,
Terminal.app, and tmux).
Use agentpulse --here to run it in the current terminal instead (handy over SSH
or inside a tmux pane).
Agent Pulse is a real-time terminal dashboard that monitors your GitHub Copilot CLI sessions, agents, and activity. Built with Python, Textual, and Rich, it gives you full observability into your AI-powered development workflow.
| Feature | Description |
|---|---|
| 🖥️ Live Session Tracking | Real-time monitoring of active Copilot CLI terminal sessions |
| 🤖 Agent Monitoring | Track 15+ agent types: task, explore, general-purpose, rubber-duck, code-review, and custom agents |
| 📊 14-Day Trend Analysis | Sparklines, daily breakdowns, gradient bar charts, and trend arrows |
| 🔥 24h Activity Heatmap | Hourly session density visualization with ░▒▓█ blocks |
| 🚀 Agent Velocity | Agents-per-hour metric with peak concurrent tracking |
| 📱 Responsive Layout | 3-tier adaptive UI: full → compact → micro dashboard |
| ⚡ Real-time Updates | Configurable refresh rate with pulse wave animation |
| 💾 Persistent History | Daily stats saved to ~/.copilot/agent-pulse/ |
| 🔒 DB Lock Resilience | Graceful handling when the session store is busy |
| 🎨 Polished Aesthetic | Clean color palette with gradient effects and heartbeat animations |
# Clone the repo
git clone https://github.com/DUBSOpenHub/copilot-cli-agent-pulse.git
cd copilot-cli-agent-pulse
# Create a virtual environment and install dependencies
python3 -m venv .venv
.venv/bin/pip install -r requirements.txt
# Launch the dashboard
agentpulseThat's it. The dashboard auto-detects your Copilot CLI sessions and starts monitoring.
Add these aliases to your ~/.zshrc or ~/.bashrc (fish users: ~/.config/fish/config.fish):
alias agentpulse='~/copilot-cli-agent-pulse/start.sh'
alias agentdashboard='~/copilot-cli-agent-pulse/start.sh'
alias agentpulse-here='~/copilot-cli-agent-pulse/start.sh --here'Then just type agentpulse or agentdashboard from anywhere — the live dashboard automatically opens in a new terminal window so it never blocks your current session. The launcher auto-detects Ghostty, iTerm, Kitty, WezTerm, Alacritty, Warp, Terminal.app, tmux, gnome-terminal, and xterm.
Use agentpulse --here (or the agentpulse-here alias) to run the dashboard in the current terminal instead — ideal for SSH sessions or tmux panes.
| Mode | Command | Description |
|---|---|---|
| Live | python agent_pulse.py |
Launch the live Textual dashboard (default) |
| Export | python agent_pulse.py --export |
JSON export to stdout |
| No Splash | python agent_pulse.py --no-splash |
Skip boot animation |
| Flag | Description |
|---|---|
--here |
Run in the current terminal (one-off; doesn't change your saved preference) |
--new-window |
Open a new window (one-off; doesn't change your saved preference) |
--reconfigure |
Forget the saved launch-mode preference and ask again on next run |
💡 First run: Agent Pulse asks whether you prefer a new window or in-place, then remembers your choice in
~/.config/agent-pulse/launcher.conf. Explicit flags always override the saved preference.
--export, -e Export JSON to stdout
--no-splash Skip boot animation
--version, -v Show version number
┌──────────────────┐ ┌──────────────────┐ ┌──────────────────┐
│ Process Scanner │ │ Session Store │ │ Event Parser │
│ (ps aux) │────▶│ (SQLite DB) │────▶│ (events.jsonl) │
└──────────────────┘ └──────────────────┘ └──────────────────┘
│ │ │
▼ ▼ ▼
┌─────────────────────────────────────────────────────────────────────┐
│ collect_all_stats() │
│ Merges: procs, sessions, DB stats, agents, skills, history │
└─────────────────────────────────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────────────┐
│ Adaptive Layout Engine │
│ Full (≥100w, ≥40h) → Compact (≥80w) → Micro (<80w or <24h) │
└─────────────────────────────────────────────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────────────┐
│ Textual App (TUI) │
│ Banner │ Metrics │ Sessions │ Agents │ Health │ Models │
└─────────────────────────────────────────────────────────────────────┘
| Source | Path | What It Provides |
|---|---|---|
| Process table | ps aux |
Active Copilot CLI processes (PID, CPU, MEM) |
| Session store | ~/.copilot/session-store.db |
Total/daily/weekly/monthly session counts, turn counts |
| Session state | ~/.copilot/session-state/ |
Active sessions, lock files, event streams |
| Agent registry | ~/.copilot/agents/ |
Installed agent definitions |
| History cache | ~/.copilot/agent-pulse/history.json |
Persistent daily statistics |
- ASCII art title with animated pulse wave (
▁▂▃▄▅▆▇█▇▆▅▄▃▂) - Heartbeat animation, uptime counter, live clock
- Dynamic border: green when active, cyan when idle
- Active sessions, processes, and 24h agent count
- Gradient bar charts (green → yellow → red)
- Peak concurrent, agent velocity, last agent launched
- 24h activity heatmap with
░▒▓█density blocks - Real-time session and launch sparklines (rolling 4-min window)
- 7-day daily breakdown with gradient bars
- 14-day sparklines with trend arrows (↑↓→)
- Session and agent counts side-by-side
- Stacked distribution bar by agent type
- Colored badges for 15+ agent types
- Gradient share bars per type
- Skill invocation tracking
- Live table of running Copilot CLI sessions
- Session ID, PID, agent type, status, and runtime
- Auto-discovered agent registry from
~/.copilot/agents/ - Icon, name, and description for each agent
- Health score gauge (0–100) with status label (Excellent / Good / Warning / Critical)
- 24h success rate and error count
- Pulsing monitor indicator
- 24h breakdown of AI models used across agent launches
- Per-model count, percentage, and gradient bar
- 24h token consumption with estimated cost
- Hourly token sparkline
copilot-cli-agent-pulse/
├── agent_pulse.py # Main dashboard application
├── agent_pulse.tcss # Textual CSS stylesheet
├── pyproject.toml # Python packaging + entry point
├── requirements.txt # Python dependencies (rich, textual)
├── start.sh # Launcher (auto-detects terminal emulator; --here for in-place)
├── quickstart.sh # One-command installer
├── site/ # Showcase website (GitHub Pages)
│ └── index.html
├── assets/ # Screenshots and images
├── experimental/
│ └── ink/ # React/Ink TUI (experimental)
│ ├── src/
│ └── package.json
├── .github/
│ ├── copilot-instructions.md
│ └── workflows/
│ ├── ci.yml
│ └── pages.yml
├── AGENTS.md
├── CONTRIBUTING.md
├── LICENSE
├── README.md
└── SECURITY.md
Requirements: Python 3.10+ and an active GitHub Copilot CLI installation.
An alternative dashboard implementation using React and Ink lives in experimental/ink/. This is a separate TUI with its own component architecture.
cd experimental/ink
npm install
npm startNote: The Python implementation is the primary, production-ready version. The Ink version is experimental.
See CONTRIBUTING.md for guidelines.
See SECURITY.md for our security policy.
MIT — built with ❤️ for the GitHub Copilot CLI community.
Built with ❤️ for the GitHub Copilot CLI community by @DUBSOpenHub.