AI Report to Me

Let your coding agent report to you.

English | 中文

aireport reads your local coding agent session logs, analyzes patterns, and generates daily reports and 90-day Wrapped summaries — complete with friction analysis, habit profiling, and a "Vibe Coder" personality type.

It works with Claude Code, Gemini CLI, OpenCode, and Codex. All data stays on your machine.

Features

• Daily Reports — Structured summaries of what you built, what broke, and what you learned. Includes "Experience Slices": distilled problem-solving stories with transferable insights.

  Feishu	Slack

• 90-Day Wrapped — A Spotify Wrapped-style retrospective with session stats, hourly heatmaps, top projects, and a "Vibe Coder" personality classification (e.g. Night Ghost, Flash Raider, Refactor Addict).

  

• Friction Detection — Automatically identifies retry loops, error cascades, and direction switches in your sessions.

• Startup Reminders — Claude Code's SessionStart hook nudges you to generate yesterday's report. OpenCode gets a custom ai_report_check tool you can invoke in chat.

• Webhook Push — Send reports to Slack, Discord, Feishu, DingTalk, WeCom, or Microsoft Teams.

• Multi-Language — Reports generated in English, Chinese, Japanese, Korean, or Russian.

Quick Start

Highly Recommended: paste this into your coding agent

(Claude Code, Gemini CLI, OpenCode, or Codex) and let it handle everything:

Install and configure ai-report-to-me by following the instructions here:
curl -s https://raw.githubusercontent.com/FakeHank/ai-report-to-me/main/docs/agent-install.md

The agent will read the guide, ask for your preferences (language, webhooks), install the package, and configure hooks automatically.

Or install manually:

curl -fsSL https://raw.githubusercontent.com/FakeHank/ai-report-to-me/main/scripts/install.sh | bash

Once installed, use the built-in commands directly in your coding agent session:

CLI Tool	Daily Report	90-Day Wrapped
Claude Code	/dayreport	/qtreport
Gemini CLI	/dayreport	/qtreport
OpenCode	/dayreport	/qtreport
Codex	$dayreport	$qtreport

How It Works

┌─────────────┐     ┌───────────────┐     ┌────────────┐     ┌───────────┐
│   Adapters   │────▶│  Aggregator   │────▶│  Analyzers  │────▶│  Renderer  │
│              │     │               │     │             │     │           │
│ Claude Code  │     │ Group by day  │     │ Friction    │     │ Markdown  │
│ Gemini CLI   │     │ Project stats │     │ Habits      │     │ Vibe Card │
│ OpenCode     │     │ Token usage   │     │ Vibe Type   │     │ Webhooks  │
│ Codex        │     │ Tool calls    │     │ Suggestions │     │           │
└─────────────┘     └───────────────┘     └────────────┘     └───────────┘

1. Adapters parse session logs from each CLI tool into a normalized format
2. Aggregator groups sessions by date and computes per-project breakdowns
3. Analyzers detect friction patterns, classify coding habits, and extract personality signals
4. Renderer produces Markdown reports (with LLM-generated narrative) and PNG vibe cards

Reports are saved to ~/.ai-report/reports/ with embedded metadata for incremental regeneration — only new sessions trigger a rebuild.

Supported CLIs

CLI Tool	Session Source	Hook Support	Notes
Claude Code	~/.claude/projects/	Full	SessionStart hook (startup check)
Gemini CLI	~/.gemini/tmp/	None	Session data read directly
OpenCode	~/.local/share/opencode/opencode.db	Partial	Plugin with custom ai_report_check tool
Codex	~/.codex/sessions/	None	Session data read directly

CLI Reference

aireport install           # interactive setup (select sources, language, webhooks)
aireport daily             # generate daily report(s)
aireport daily --date 2026-02-28   # generate for a specific date
aireport wrapped           # generate 90-day Wrapped
aireport wrapped --days 30         # custom period
aireport status            # show installation status and session stats
aireport config            # manage configuration
aireport uninstall         # remove hooks and clean up

Configuration

Configuration lives at ~/.ai-report/config.json and is created during aireport install:

{
  "output_lang": "en",           // en | zh | ja | ko | ru
  "sources": ["claude-code"],    // which CLIs to read from
  "daily_reminder": true,        // show startup reminder in Claude Code
  "backfill_limit": 7,           // days to look back for pending reports
  "privacy_mode": false,         // redact file paths and content in reports
  "wrapped_days": 90,            // default period for Wrapped
  "webhooks": {
    "slack_url": "",             // Slack Incoming Webhook URL
    "discord_url": "",           // Discord Webhook URL
    "feishu_url": "",            // Feishu/Lark Bot Webhook URL
    "dingtalk_url": "",          // DingTalk Robot Webhook URL
    "wecom_url": "",             // WeCom/WeChat Work Bot URL
    "teams_url": ""              // Microsoft Teams Webhook URL
  }
}

Webhooks

When a webhook URL is configured, reports are automatically pushed after generation. Each platform gets a native-formatted message:

Platform	Format
Slack	Block Kit
Discord	Embeds
Feishu / Lark	Interactive Card
DingTalk	ActionCard
WeCom	Markdown
Microsoft Teams	Adaptive Card

Data & Privacy

All session data is read from your local filesystem and never leaves your machine. Reports are stored locally at ~/.ai-report/. The optional privacy_mode setting redacts file paths and code content from report data before it reaches the LLM.

Webhooks send the generated report text (not raw session data) to your configured endpoints.

Contributing

git clone https://github.com/FakeHank/ai-report-to-me.git
cd ai-report-to-me
pnpm install
pnpm dev -- status         # run any CLI command in dev mode
pnpm test                  # run tests
pnpm typecheck             # type-check without emitting

The codebase is ESM-only TypeScript (Node 20+, strict mode). Tests use Vitest.

License

MIT