AI-Agent for SRE & DevOps — Built for production, not playgrounds.
Logs too big? Errors hidden? Shello handles it.
Shello is a production-grade AI terminal assistant that executes commands, manages giant outputs, and extracts critical errors without blowing up your token budget.
Explore Features • Quick Start • Report Bug
| Feature | Standard AI Tools | 🐚 Shello CLI |
|---|---|---|
| Giant Log Outputs (50k+ lines) | Crash, refuse, or burn $10 in tokens | Semantic Truncation (keeps errors visible) |
| Command Safety & Trust | Blind execution without safety guardrails | Allowlist & Denylist safety gates with user approval |
| Verification & Context | Loose context / terminal memory limit | Persistent LRU Cache for full AI memory |
| Massive JSON Payloads | Dump raw text into the prompt | JSON Intelligence (navigates via JSON Query (jq) paths) |
| Model Flexibility | Locked into one model/provider | Runtime provider switching (OpenAI ⇄ Bedrock) |
Run the command for your operating system:
Windows (PowerShell - No Admin needed):
# Create folder, download executable, and add to User PATH
New-Item -ItemType Directory -Force -Path "$env:USERPROFILE\.shello_cli" | Out-Null
Invoke-WebRequest -Uri "https://github.com/om-mapari/shello-cli/releases/latest/download/shello.exe" -OutFile "$env:USERPROFILE\.shello_cli\shello.exe"
[Environment]::SetEnvironmentVariable("PATH", $env:PATH + ";$env:USERPROFILE\.shello_cli", "User")macOS:
curl -L https://github.com/om-mapari/shello-cli/releases/latest/download/shello-macos -o /tmp/shello && sudo mv /tmp/shello /usr/local/bin/shello && sudo chmod +x /usr/local/bin/shelloLinux:
curl -L https://github.com/om-mapari/shello-cli/releases/latest/download/shello -o /tmp/shello && sudo mv /tmp/shello /usr/local/bin/shello && sudo chmod +x /usr/local/bin/shelloVerify that the CLI is installed correctly:
shello --versionConfigure your API provider (first time setup):
shello setupStart the interactive session:
shelloWhen logs are too large, Shello handles it instantly without burning tokens. It strips the noise and shows you the actual failure context.
🌊 user [~/k8s-cluster]
──└─⟩ why is my pod crashing?
🐚 Shello
$ kubectl get pods
NAME READY STATUS RESTARTS AGE
api-deployment-abc123 0/1 CrashLoopBackOff 5 3m
$ kubectl logs api-deployment-abc123
[... Truncating 2,000 lines of normal startup logs ...]
Error: ECONNREFUSED connect to database:5432
at TCPConnectWrap.afterConnect [as oncomplete]
🔍 Root Cause Found:
Your pod cannot connect to the database on port 5432.
Check your database service and the connection string in your deployment config.
When commands produce thousands of lines (e.g., builds, package installs), Shello truncates the display to keep the user's terminal clean. Behind the scenes, the full log is cached locally so the AI can query specific log segments in the background to diagnose issues.
🌊 user [~/projects/myapp]
──└─⟩ npm install
🐚 Shello
$ npm install
[... Truncated 5,000 lines of npm logs. Full output cached as cmd_002 ...]
✓ Added 1,247 packages in 45s.
# If a package fails to install, the AI automatically calls:
# get_cached_output(cache_id="cmd_002", lines="-100")
# to inspect the exact failure stack trace in the background.
Shello asks for permission before running potentially unsafe commands, keeping your environment protected.
🌊 user [~/project]
──└─⟩ delete the node modules folder
🐚 Shello
┌─[💻 user@computer]─[~/project]
└─$ rm -rf node_modules
⚠️ COMMAND APPROVAL REQUIRED
Choose an action:
❯ [A] Approve command
[D] Deny command
[C] Provide custom feedback to AI
- 🛡️ Command Trust & Safety A robust safety system checks commands before execution using a custom allowlist, denylist, and AI-driven safety flags. Skip verification entirely using YOLO mode.
- 📉 Semantic Output Truncation No more terminal floods or wasted tokens. Shello compresses progress bars, removes column padding, and ensures errors/warnings remain visible in massive log streams.
- 💾 Progressive Disclosure Cache
Shello caches full terminal output locally (
cmd_001,cmd_002). The AI dynamically retrieves only the relevant segments it needs to troubleshoot. - 🧠 JSON Intelligence Analyzes giant JSON blobs by querying specific paths using JSON Query Language (jq) instead of dumping the entire raw JSON payload to the prompt.
- 🔄 Multi-Model & Provider Switching Seamlessly switch between AI providers (OpenAI-compatible APIs, OpenRouter, AWS Bedrock) and models (Claude, Nova, GPT) dynamically during a session.
- 🐳 Production-Ready Tailored specifically for SRE & DevOps workflows debugging Cloud, Kubernetes, Docker, and Linux systems.
Configure your preferred AI provider (OpenAI-compatible APIs or AWS Bedrock) and default models by running the interactive setup wizard:
shello setupThe wizard automatically creates a global configuration file at ~/.shello_cli/user-settings.yml with helpful inline comments.
- Manual Tweaks: For detailed YAML settings, custom token limits, and advanced provider configuration, see the Development Setup Guide.
- Credentials: You can also set credentials via environment variables (
OPENAI_API_KEY,AWS_REGION, etc.). See Development Setup Guide for details.
While in a chat session:
/new— Start a fresh conversation./switch— Dynamically switch between AI providers and models without losing chat history./help— List available chat commands./quit//exit— Exit Shello.
Global CLI commands:
shello setup— Run interactive configuration wizard.shello config— View current configuration settings.shello config --edit— Edit settings in your default editor.shello --version— Display version.
Shello protects your system from executing dangerous commands via a built-in trust verification system:
- Safe commands (like
ls,git diff,grep) execute instantly. - Potentially unsafe commands trigger an interactive approval dialog prompting you to Approve, Deny, or provide Custom Feedback to the AI.
- Highly destructive commands (like
rm -rf /) are blocked by a strict denylist.
To customize the allowlist/denylist patterns, configure YOLO mode, or toggle AI safety gates, see the Command Trust and Safety Guide.
For developers who want the details:
- Hybrid execution model - Direct shell for instant commands, AI routing for analysis
- Formal correctness properties - 8 properties tested via property-based testing (Hypothesis)
- Intelligent truncation - Type detector, semantic classifier, progress bar compressor keep errors visible
- Persistent LRU cache - Sequential IDs (cmd_001, cmd_002...), 100MB limit, per-conversation
- Streaming architecture - Real-time output for you, processed summary for AI (no token waste)
- Zero data loss - Full output always cached, grab any section for deeper debugging
- Modular design - Clean separation: cache → detect → compress → truncate → analyze
- Token optimization - Strips column padding, compresses progress bars (2-3x fewer tokens)
If you want to set up Shello CLI for local development, run tests, or build the standalone executable from source, please refer to the Development Setup Guide for full instructions.
Contributions welcome! Fork the repo, make a feature branch, and send a PR.
Check CONTRIBUTING.md for guidelines.
Om Mapari
- GitHub: @om-mapari
- Project: Shello CLI
MIT License - see LICENSE
Built with:
- Click - CLI framework
- Rich - Terminal UI
- OpenAI - AI capabilities
- AWS Bedrock - Multi-model support
Made with ☕ and 🐚 by Om Mapari