A local desktop app that helps you decide whether an X post is worth replying to and drafts one sharp reply if it is.
Paste a public X link or raw post text. The app imports the post, reasons through what it is actually saying, suggests one reply, and gives you a clear Comment or Skip verdict.
- Import a public X post from a link or paste raw text
- Get a single reply recommendation instead of a pile of variations
- Choose between Quick Local, Focused AI, Claude, and Ollama modes
- Keep the app local-only on
127.0.0.1
- Node.js 18 or newer
- No npm dependencies are required
- Optional API keys if you want hosted model modes
- Optional Ollama runtime if you want local model mode
Double-click launch.bat.
The launcher finds a free port between 8785 and 8795, starts the local server, waits for /api/health, and opens the app in your browser automatically.
node server.jsThe manual server defaults to http://127.0.0.1:8765.
You can also choose a port explicitly:
node server.js 8785Copy .env.example to .env if you want server-side defaults.
OPENAI_API_KEY=
ANTHROPIC_API_KEY=
PORT=8765
X_REPLY_MODEL=gpt-5
X_REPLY_REASONING_EFFORT=high
CLAUDE_MODEL=claude-sonnet-4-5
OLLAMA_BASE_URL=http://127.0.0.1:11434
OLLAMA_MODEL=| Mode | What it uses | API key needed? |
|---|---|---|
| Quick Local | Built-in keyword and tone analysis | No |
| Focused AI | OpenAI two-pass reasoning | Yes |
| Claude | Anthropic two-pass reasoning | Yes |
| Ollama | Local Ollama runtime | No, but Ollama must be running |
index.html- single-page UIstyles.css- app stylingapp.js- frontend logic and local reasoning engineserver.js- local Node.js server and API integrationslaunch.bat- Windows launcher
- This is a manual copilot, not an auto-reply bot
- The server binds to
127.0.0.1only - Browser-entered API keys are stored in localStorage in that browser profile
- Server-side API keys can be supplied through environment variables instead
- Public post import uses the X oEmbed endpoint and only works for public posts
npm run checkMIT. See LICENSE.