Agent Mission Control banner

Monitor and analyze your AI coding agents

Track live agent activity, review session history, and chat with your usage analytics to spot trends, token hotspots, model shifts, and recurring failures.

Loading…
Dashboard preview

Keep an eye on your Copilot CLI and Copilot App agentic coding sessions without breaking focus. Every tool call, sub-agent dispatch, and permission prompt becomes a live signal on a second monitor — so you spot stalls, surprises, and "needs attention" the moment they happen, instead of scrolling back through terminal output.

⬇ Download Now ⌥ View on GitHub
Tauri 2 Phaser 4 Rust + TypeScript Mac · Win · Linux

🛰️ Mission Control

Each Copilot tool category becomes a sector on the mission ring around the central outpost. Each sector lights up as work flows through it.

⚒️

Edits

File writes and apply_patch calls. Lights up brightest when edits are flying.

apply_patchwrite
📚

Reads

Reads, search, and code navigation. Usually the busiest sector in any session.

viewrggrep
⚙️

Commands

Shell commands the agent runs — builds, tests, lints, anything that touches a process.

bashshell
📡

Web / Docs

External lookups — web fetches, docs research, anything reaching beyond your machine.

web_fetchweb_search
🛡️

Sub-Agents

Sub-agent delegations. Watch the main agent dispatch specialized workers.

taskagent
📖

Skills

Skill invocations — the agent pulling in pre-built playbooks for specialized work.

skill
🪶

MCP

MCP tool calls — external servers extending the agent's capabilities.

mcp.*
👑

Intent / Alerts

Permission prompts and alerts. When this sector lights up, the agent needs a decision.

promptalert

⚡ How It Works

01

Launch the app

Download a release for macOS, Windows, or Linux. Open it once and the window remembers its size and position for next time. No setup, no accounts.

02

Run Copilot CLI in your terminal

Use Copilot CLI like you normally would. Agent Mission Control reads the session state Copilot already writes under ~/.copilot/session-state/ — there's no extra config.

03

Watch Mission Control react

Tool calls light up sectors within ~300 ms of happening. Use the replay scrubber to rewind. Click the 👁 button to hide the side panels and put Mission Control front-and-center on a second monitor.

04

Inspect tools and turns

Open the selected-session inspector to filter recent MCP, skill, sub-agent, and failed tool calls. Switch to turn mode to see the agent's recent work grouped by model turn, with only sanitized details crossing the app bridge.

05

Review global History

Switch to the History route for aggregate analytics across all scanned sessions: 24-hour and 7-day activity charts, model mix, event mix, top tools, recent sessions, and sanitized failure history. Prompts, raw arguments, command output, file paths, and diffs stay out of the view.

06

Open Analytics Chat

Open Analytics Chat to ask natural-language questions about your Copilot CLI history. It answers from local derived metrics and can include charts, tables, trends, and recommendations for patterns like token hotspots, model shifts, repeated tool failures, and MCP server usage.

🛠️ How This Was Made

Built with GitHub Copilot CLI and GitHub Copilot App, which scaffolded the Tauri 2 + Phaser 4 + TypeScript shell and then iteratively designed the mission layout, the privacy-safe AgentProvider trait, and the replay timeline.

🙏 Credits

🦀

Tauri 2

Tauri 2 — the Rust-backed desktop shell.

🎮

Phaser 4

Phaser 4 — the WebGL scene that draws Mission Control.

🤖

Copilot CLI

GitHub Copilot CLI — AI-powered command-line assistance.

🧭

GitHub Copilot App

GitHub Copilot App — orchestrate AI coding agents from a single desktop app.