Stop copy/pasting errors. Let your terminal talk to AI.
Install · Usage · How it Works · Contributing
errpipe is an interactive shell wrapper built in Go that monitors your commands. When a command fails, it automatically captures the error output and sends it to your preferred LLM for immediate analysis and fixes.
Instead of switching to your browser manually, errpipe brings the fix to you, either directly in your terminal or by automating your browser.
Windows (PowerShell)
irm https://diamondosas.github.io/errpipe/docs/install/install.ps1 | iexmacOS / Linux
curl -fsSL diamondosas.github.io/errpipe/docs/install/install.ps1 | shFrom Source (Go required)
Ensure you have Go installed (1.21+ recommended).
git clone https://github.com/diamondosas/errpipe
cd errpipe
go mod tidy
go build -o errpipeAdd the resulting binary to your system PATH.
Before your first session, run the interactive setup:
errpipe --initFollow the prompts to choose your provider (Gemini, Claude, or ChatGPT) and your preferred mode.
Run errpipe to start the interactive session.
errpipeOnce inside the errpipe shell, run your commands as usual:
[EP] C:\projects\myapp> go build
# If it fails, errpipe captures the stderr and triggers the AI--help: Show help message.--init: Initialize/Reconfigure the application.exit: Leave theerrpipeshell.
errpipe supports three primary ways to interact with AI:
| Provider | Status | Inline (Streaming) | CLI Mode | Web Mode |
|---|---|---|---|---|
| Google Gemini | ✅ Supported | ✅ Yes | ✅ Yes | ✅ Yes |
| Anthropic Claude | ✅ Supported | ✅ Yes | ✅ Yes | ✅ Yes |
| OpenAI ChatGPT | ✅ Supported | ✅ Yes | ✅ Yes | ✅ Yes |
- Inline CLI Mode: Streams AI responses directly into your terminal. This is the fastest method but requires an API Key from the provider.
- CLI Mode: Interacts with the official CLI tools of the providers installed on your system.
- Web Mode: Automatically detects your browser, opens the provider's chat page, and types the error message for you using browser automation.
- REPL:
errpipeacts as a thin wrapper around your default shell (cmd/powershellon Windows,sh/bashon Linux/macOS). - Monitor: It pipes
stdoutandstderrto your terminal while also capturingstderrin a buffer. - Trigger: If a command returns a non-zero exit code,
errpipesends the capturedstderrto the configured AI service. - Analysis: Depending on your mode, it will either stream the fix directly or automate your environment to get you the answer.
errpipe runs locally and only sends data to the LLM when a command fails. Be mindful of sensitive information in your error logs before sending them to public AI models.
PRs are welcome.
- Fork the repo
- Create your feature branch (
git checkout -b feature/amazing) - Commit your changes (
git commit -m 'Add amazing feature') - Push to the branch (
git push origin feature/amazing) - Open a Pull Request
MIT.
