The bug stops here
DebugAI sits in VS Code, reads the stack trace, pulls the files that actually matter, and hands you a fix you can apply in one click. Built for the errors your copilot shrugs at.
Nothing to paste.
[run]A command fails
Press Ctrl+Shift+P and run DebugAI: Analyze Terminal Error. It reads the error from your terminal, so there is nothing to copy.
[read]It reads your codebase
Traced files come first. Then the related code your trace doesn't mention, retrieved by embeddings across the repo. The model sees what a senior teammate would open.
[explain]Root cause, plain language
If this error has hit your project before, the fix that worked leads the answer. Haiku handles the easy ones. On Pro and Team, the ugly ones go to Sonnet.
[apply]One click, native diff
You review it like a PR, because it is presented like one. Nothing writes to your files without your click.
You've read this stack trace before. So has it.
Every debug session leaves a signature. When the same error resurfaces in your project, DebugAI leads with the fix that was confirmed to work. Project-scoped, never shared across accounts.
Your agent can hand us the error itself.
DebugAI ships as a standalone MCP server, and one command sets it up. It signs you in through your browser, then writes the config for every MCP client on your machine (Claude Desktop, Claude Code, Cursor, Zed, Windsurf, Cline, Gemini CLI). Each one gets adebug_error tool: the agent sends the stack trace, the same engine behind the extension sends back the root cause and ranked fixes. No key to paste. No VS Code required.
npx -y @debugai/mcp setupsetup for every client, on npm ↗$12. Not $20. You noticed too.
- 10 debugs a day
- fix applied in your editor, with Undo
- full codebase context
- no card, no trial clock
10/day covers most Tuesdays. Resets at midnight UTC.
- 1,000 debugs a month, soft cap
- history kept for good (Free keeps 7 days)
- Sonnet routing on hard errors
- error memory
- 14-day trial, opt-in, no card
Soft cap means we tell you at 1,000 and do not stop you there. Fair use applies above it. Annual: $120 (two months free).
Fair questions.
Is my code sent to your servers?
In two cases. A debug sends the error message and the code around it. Indexing uploads the files it indexes, which are split into chunks, embedded and stored under your account, and used only to answer your own debugs. Other users never see them.
How is this different from ChatGPT?
ChatGPT only knows what you paste. DebugAI indexes your project (up to 200 files on Free): file structure, imports, call chains. It traces errors across files automatically. No architecture explanations needed.
What languages and frameworks does DebugAI support?
Python (FastAPI, Django, Flask, Streamlit, PyTorch) and JavaScript/TypeScript (React, Next.js, Express, Node.js, Vite). Java and C# projects are indexed and analysed as well. More added based on user feedback.
Is DebugAI actually free?
Yes. 10 debug sessions per day on the free tier, forever. No credit card required.
Will DebugAI overwrite my code without asking?
Never. Every fix shows a full diff before anything is applied. After you apply one, Undo stays on it until your next analysis, and Ctrl+Z works too.
Does DebugAI work with Cursor or Windsurf?
Yes. DebugAI is a standard VS Code extension and works in any VS Code-compatible editor including Cursor and Windsurf.
Can my AI agent use DebugAI without VS Code?
Yes. Run npx -y @debugai/mcp setup. It signs you in through your browser and writes the config for every MCP client it finds: Claude Desktop, Claude Code, Cursor, Zed, Windsurf, Cline, Gemini CLI. Each one gets the debug_error tool. Nothing to paste.
How long does DebugAI setup take?
About 30 seconds. Install the extension, run "DebugAI: Sign In" from the command palette, and confirm a short code in your browser. There is no API key to find or copy. For agents outside VS Code it is one command: npx -y @debugai/mcp setup.