🚀 Send to AI — Copy Codebase to ChatGPT, Claude & Gemini

The only VS Code extension that both bundles your code for any AI chat and applies the reply back. No API key. No telemetry. Nothing leaves your machine except what you paste.

Panel at a glance

That's the whole loop, inside your editor:

1. 🐛 Bundle Errors — grab every file VS Code is flagging + the exact error messages.
2. Paste into ChatGPT, Claude, Gemini — whatever you already pay for (or don't).
3. 📥 Apply AI Reply — paste the answer back; it writes the edits to your files, reviewable and undoable.

No web tool that's free of an API key does this — because no web tool can see your Problems panel or write to your files. That's the editor-native moat.

✨ What makes it different

🐛 Bundle Errors — task-aware by construction

One click bundles only the files VS Code is reporting errors/warnings on, with the exact diagnostics (src/api.ts:42:7 [ERROR] Property 'id' does not exist…) and a ready-to-paste "please fix these" prompt. Web bundlers can't read your Problems panel. This can.

It captures a snapshot of the Problems panel as it is right now. If the errors look stale or wrong, save the file or run "Developer: Reload Window" / "Restart Language Server" so the diagnostics refresh first.

📥 Apply AI Reply — the round trip

Copy the AI's response (code blocks labelled with file paths), run the command, and pick which edits to apply. Every write is:

• ✅ Reviewable — shows up in the editor and in git diff
• ✅ Undoable — a single Ctrl+Z
• ✅ Safe — you choose each file; it creates new ones or replaces existing ones

No more hand-copying snippets back block by block.

Tip: copy the AI's whole reply, not the chat's per-block "copy code" button — the file path and the ``` fences must be on the clipboard for the edits to be detected.

🎯 Relevance ranking — keep what matters (Pro · part of Fit to AI Window)

When a bundle must be trimmed to fit a token window, files are ranked by relevance to your actual task — filename/path/content overlap with your prompt, plus structural priors that keep entry points, READMEs and config and demote tests, generated code and lockfiles. Dumb bundlers cut the biggest files; this cuts the least relevant ones.

🔗 Dependency-graph awareness — connected, not orphaned (Pro · part of Fit to AI Window)

A file imported by a relevant file inherits part of that relevance, so the local dependencies of what you're working on survive the cut alongside it. Follows relative import/require/from edges in JS/TS and Python. You get complete, connected context — not orphaned snippets.

💰 Free + Pro ($14.99 one-time)

Works with any AI you already use — it's a companion to your existing chat subscription (or a free one), not a replacement for it. No monthly fee, no API key.

🆓 Free tier

• Project bundling — bundles the 50 most relevant files of any project (never a paywall: you always get a usable bundle, ranked by relevance to your task)
• 🐛 Bundle Errors + 📥 Apply AI Reply — the full round-trip loop
• All 4 output formats (Standard, Claude XML, Compact, Minimal)
• Visual file picker with live per-file token estimates
• Open Tabs mode, Project Notes (persistent context), token + cost estimator
• Context-window bars for ChatGPT / Claude / Gemini
• .gitignore + .sendtoaiignore aware auto-ignore

💎 Pro tier ($14.99 lifetime)

• Unlimited files per bundle
• 🔀 Git diff mode — bundle only changed/staged files for AI code review
• 🎯 Fit to AI Window — auto-trim to your AI's context using relevance ranking + dependency-graph awareness (keeps what matters, not the biggest files)
• 📋 Named presets (up to 10) — one-click reload of "backend only" etc.
• Use it on up to 5 machines — run Deactivate License on This Machine to move a slot
• All future Pro features included

🚀 Upgrade — $14.99 lifetime · plus VAT where applicable · No subscription. 30-day money-back guarantee.

Working in a team? The Team pack ($39 once, 5 developers, 25 activations on one key) is on sendtoai.dev.

📦 Quick start

1. Install the extension from the VS Code Marketplace or, for Cursor, Windsurf and VSCodium, from Open VSX — the SendToAI icon appears in the activity bar.
2. Open the panel and use the buttons, or right-click a file/folder → Send to AI.
3. Paste into your favourite AI chat — then copy its reply and hit 📥 Apply AI Reply.

Typical loop

🐛 Bundle Errors  →  paste into Claude/ChatGPT  →  copy reply  →  📥 Apply AI Reply  →  Ctrl+Z if you don't like it

🎮 Usage

Sidebar panel

The fastest path — open the SendToAI view and click:

• 🐛 Bundle Errors — bundle everything with a problem + the messages
• 📥 Apply AI Reply — apply the AI's code blocks back to your files
• Bundle project, pick output format, set a token window, estimate cost

Keyboard shortcuts

• Ctrl+Alt+B (Cmd+Alt+B): open the SendToAI panel, ready to bundle the project
• Ctrl+Alt+A (Cmd+Alt+A): copy the current file (or selection) to the clipboard with its path

Command Palette

Ctrl+Shift+P → type "Send to AI" → pick any command (including Bundle Errors / Apply AI Response).

Pro license commands: Enter License Key, Deactivate License on This Machine (frees one of your 5 slots before moving to a new PC), and Show License Activity (a timestamped activation log to send support if anything goes wrong).

⚙️ Configuration

File > Preferences > Settings → search "Send to AI":

{
    "sendtoai.licenseKey": "XXXXXXXX-XXXX-XXXX-XXXX-XXXXXXXXXXXX",
    "sendtoai.autoOpenAI": false
}

Custom ignore patterns

Create a .sendtoaiignore in your project root — same syntax as .gitignore:

# .sendtoaiignore
secrets/**
*.generated.ts
docs/

🤝 Works with every AI

• ChatGPT (GPT-5 family and newer)
• Claude (Haiku, Sonnet, Opus)
• Google Gemini (Pro, Ultra)
• Local models (Ollama, LM Studio)

Output formats include an XML mode tuned for Claude and structured analysis, plus Compact/Minimal for tight token budgets.

🔒 Privacy & security

• ✅ Your code stays on your machine — bundling is 100% local; the only place it goes is the clipboard, and from there to whichever AI you paste it into
• ✅ No telemetry, no analytics — the free tier makes zero network calls
• ✅ Pro license check is the one exception — with a key set, the extension verifies it against api.lemonsqueezy.com (sends the key plus an activation name made of your computer's hostname and a random per-install id, cached for 24 hours). Nothing about your code is included
• ✅ Secrets are never bundled — .env*, private keys and certificates (.pem, .key, .p12…), .npmrc/.netrc, cloud credential files and Terraform state are always skipped, symlinks are never followed, and files over 1 MB are left out
• ✅ 0 known vulnerabilities — dependencies patched and pinned (npm audit clean)
• ✅ Open source — audit it yourself

🎯 Perfect for

• 🐛 Fixing a wall of compiler/linter errors in one paste
• 📊 AI code reviews across multiple connected files
• 🔄 Refactoring with full, dependency-aware context
• 🧪 Generating tests from existing code
• 💡 Architecture discussions grounded in the real codebase

📞 Support

• 🌐 Website: sendtoai.dev
• 📝 How it compares: Send to AI vs Repomix vs gitingest (dev.to)
• 🐛 Found a bug? Open an issue
• 💬 Need help? support@sendtoai.dev
• 💳 Billing & refunds: 30-day no-questions refund — email support@sendtoai.dev with your order number. Purchases are processed by Lemon Squeezy, so your receipt and card statement say "Sold through Link, LLC"
• ⭐ Love it? Rate it 5 stars

📄 License

MIT — see LICENSE.

Bundle your errors + connected code for any AI, paste the answer back, apply it — no API key. Free tier, Pro is $14.99 once. 🔧