Analyze your AI coding assistant usage across VS Code, Cursor, GitHub Copilot for Xcode, Claude Code, Codex, OpenCode, and GitHub Copilot CLI — and optionally sync to Truxt SEI for team-wide, Gemini-graded coaching and adoption analytics.
Highlights
The extension is organized into three sections: Observe, Measure, and Improve.
Observe
| Page | What it shows |
|---|---|
| Dashboard | Practice scores with week-over-week and month-over-month trends, unlocked achievements, skill finder summary, daily activity chart with per-harness breakdown, and top workspace stats |
| Timeline | Gantt-style session timeline with per-day drill-down, session overlap detection, and a searchable list view |
| Coding Moments | Screenshot gallery from AI coding sessions with story reels, workspace filtering, and progressive image loading |
Measure
| Page | What it shows |
|---|---|
| Output | Two tabs -- Code Output (generated code volume by language and workspace) and Token Usage (model usage table with per-model token breakdown) (Token Usage temporarily hidden) |
| Burndown | Monthly token budget progress with projection (temporarily disabled) |
| Patterns | 7x24 activity heatmap and work-life balance signals |
Improve
| Page | What it shows |
|---|---|
| Anti-Patterns | Five practice score cards (Prompt Quality, Session Hygiene, Code Review, Tool Mastery, Context Management) with detailed findings, severity ratings, concrete actions, and example prompts |
| Skill Finder | AI-powered analysis of repeated prompts to discover custom skill opportunities, plus matching community skills and agents from the open-source catalog |
| Context Health | Overall context score, agentic readiness checklist, per-harness context provision breakdown, workspace context map (treemap colored by instruction quality), and AI-powered context file review |
| Rule Editor | Create, edit, and live-test detection rules as markdown with form-based or raw-source editing and AI-assisted drafting |
| Rule Playground | Interactive REPL for the rule DSL with field browser, function catalog, and metric list |
| Data Explorer | Browse request and session fields, view distributions, and run ad-hoc filters |
| Guardian | Analytics on the Prompt Guardian hook: screening funnel (accepted / rejected / ignored), weekly trend, clarity-lift score, risk flags, and prompt coverage |
Level Up
| Page | What it shows |
|---|---|
| Learning Center | Personalized quizzes and code-comparison rounds generated from your actual usage |
| Achievements | XP-based progression with Bronze, Silver, Gold, and Diamond tiers |
| Agentic SDLC | Track how you use AI across the full software-development lifecycle |
| Share | Generate a shareable stat card and export Markdown/JSON summaries |
Truxt SEI Backend mode
Sign in to your Truxt SEI deployment from the Truxt Backend tab to power the dashboards from your team's server-side analytics — Gemini-graded coaching scores, anti-pattern rules, context health, summaries, and adoption metrics for the signed-in user. Offline mode (local transcript parsing, no account needed) always remains available.
Supported Harnesses
| Harness | Default location |
|---|---|
| Local Agent | macOS: ~/Library/Application Support/Code/User/workspaceStorage/Linux: ~/.config/Code/User/workspaceStorage/Windows: %APPDATA%\Code\User\workspaceStorage\ |
| Local Agent (Insiders) | macOS: ~/Library/Application Support/Code - Insiders/User/workspaceStorage/Linux: ~/.config/Code - Insiders/User/workspaceStorage/Windows: %APPDATA%\Code - Insiders\User\workspaceStorage\ |
| Xcode Copilot Chat | ~/.config/github-copilot/xcode/ (requires sqlite3) |
| Claude | macOS/Linux: ~/.claude/projects/Windows: %USERPROFILE%\.claude\projects\ |
| Codex | macOS/Linux: ~/.codex/sessions/Windows: %USERPROFILE%\.codex\sessions\ |
| OpenCode | macOS/Linux: ~/.local/share/opencode/Windows: %USERPROFILE%\.local\share\opencode\ |
| GitHub Copilot CLI | ~/.copilot/session-state/ and ~/.copilot/history-session-state/ |
Works in Cursor (and other VS Code forks)
The extension runs in Cursor, VSCodium, and Windsurf. The dashboard and all analytics work fully. Two Copilot-tied integrations (the @aicoach chat participant and Language Model tools) require VS Code with GitHub Copilot and are automatically disabled elsewhere. For AI-powered features in Cursor, run "Truxt AI Coach: Set LLM API Key" and set truxtAiCoach.llm.provider to anthropic or openai.
Chat
In VS Code with GitHub Copilot, type @aicoach in the chat panel for conversational access to all coaching data. Slash commands /summary, /improve, /compare, and /flow give quick access to common analyses. The participant orchestrates multiple backend tools automatically to answer complex questions.
Getting Started
- Open the command palette (
Cmd+Shift+P/Ctrl+Shift+P). - Run Truxt AI Coach: Open Dashboard.
- Use the sidebar to navigate pages. Filter by workspace or harness at the bottom.
- Run Truxt AI Coach: Reload Data to re-parse after new sessions.
- Optional: sign in to Truxt SEI from the Truxt Backend tab for team analytics.
Privacy
Offline mode parses AI session transcripts entirely on your machine; nothing leaves it. Backend mode sends session data only to the Truxt SEI deployment you explicitly sign in to.