← Back to Kontasks
Powerful Pipe Commands in gvShell
Examples of powerful pipe and composition commands from gvShell documentation
Output
🔥 Powerful Pipe Commands in gvShell
1. Multi-Provider Pipelines
Chain AI providers - each processes the previous output:
gv> /pipe @claude "explain quicksort" | @openai "simplify this for beginners" gv> /pipe @ollama "draft email" | @claude "make it professional"
2. Response References
Every AI response becomes a referenceable "file":
# Reference by number gv> @claude explain monads gv> convert $1 to Python # Reference by provider name gv> show $claude # Last Claude response gv> compare $1 and $2
3. Shell Integration with $refs
Use AI responses in Unix commands:
# Word count of AI response gv> ! echo $1 | wc -w # Save response to file gv> ! cat $claude > output.txt # Pipe to system commands gv> $1.code | python # Execute code block gv> $1 | pbcopy # Copy to clipboard (macOS)
4. Compare & Benchmark
Compare responses from different providers:
# Get responses from both gv> @claude explain monads gv> @openai explain monads # Diff them gv> /diff $claude $openai gv> /diff $1 $2 --side-by-side # Benchmark all providers gv> /benchmark "explain bubble sort" PROVIDER MODEL LATENCY TOKENS COST claude sonnet 1.2s 234 $0.003 openai gpt-4o 0.9s 198 $0.002 ollama llama3 2.1s 256 $0.000
5. Export & Import
Move data in and out:
# Export to file gv> /export $1 --format md > explanation.md gv> /export session --format json > session.json # Import for context gv> /import error.log gv> @claude what's causing this error? gv> /import screenshot.png gv> @claude what's in this image?
6. Remote Multi-Agent
Execute across machines:
# Run on remote agent gv> /remote dev-vm "@claude explain this codebase" gv> /remote prod-api "@ollama --model codellama review main.go" # Broadcast to all agents gv> /broadcast "what is your current git branch?" laptop: main dev-vm: feature/auth prod-api: v2.3.1 (tag) # Delegate session to another machine gv> /delegate dev-vm with "deploy this fix"
7. Field Extraction
Access specific parts of responses:
gv> $1.code # Extract code blocks gv> $1.text # Prose only (no code) gv> $1.code[0] # First code block gv> $1.code.python # Python code blocks only gv> $1.summary # AI-generated summary
✨ Key Insight
gvShell treats AI like Unix: responses are files, providers are commands, and everything composes with pipes. This is the Unix philosophy applied to AI.
Quick Actions
Original Request
No layout configured
Details
Type
General
Status
Completed
Scope
vibetools
Tags
gvshelldocumentationpipes
Created
6 Jan 2026, 4:42 pm
Updated
6 Jan 2026, 4:42 pm
Created By
claude
Raw Data
{
"id": "f603c253-3cf4-4c88-b551-d22237d2ee53",
"type": "general",
"status": "completed",
"title": "Powerful Pipe Commands in gvShell",
"description": "Examples of powerful pipe and composition commands from gvShell documentation",
"context": {
"output": "<div style=\"background:#1e1e1e;color:#e0e0e0;padding:20px;border-radius:8px;font-family:'Consolas','Monaco',monospace;\">\n<h2 style=\"font-family:sans-serif;color:#fff;margin:0 0 16px 0;\">🔥 Powerful Pipe Commands in gvShell</h2>\n\n<div style=\"background:#2d2d2d;padding:16px;border-radius:6px;margin-bottom:16px;\">\n<h3 style=\"color:#e5c07b;margin:0 0 12px 0;font-family:sans-serif;\">1. Multi-Provider Pipelines</h3>\n<p style=\"color:#9ca3af;margin:0 0 12px 0;font-family:sans-serif;\">Chain AI providers - each processes the previous output:</p>\n<pre style=\"background:#252526;padding:12px;border-radius:4px;overflow-x:auto;margin:0;\">\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#56b6c2;\">/pipe</span> <span style=\"color:#61afef;\">@claude</span> <span style=\"color:#98c379;\">\"explain quicksort\"</span> | <span style=\"color:#61afef;\">@openai</span> <span style=\"color:#98c379;\">\"simplify this for beginners\"</span>\n\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#56b6c2;\">/pipe</span> <span style=\"color:#61afef;\">@ollama</span> <span style=\"color:#98c379;\">\"draft email\"</span> | <span style=\"color:#61afef;\">@claude</span> <span style=\"color:#98c379;\">\"make it professional\"</span>\n</pre>\n</div>\n\n<div style=\"background:#2d2d2d;padding:16px;border-radius:6px;margin-bottom:16px;\">\n<h3 style=\"color:#e5c07b;margin:0 0 12px 0;font-family:sans-serif;\">2. Response References</h3>\n<p style=\"color:#9ca3af;margin:0 0 12px 0;font-family:sans-serif;\">Every AI response becomes a referenceable \"file\":</p>\n<pre style=\"background:#252526;padding:12px;border-radius:4px;overflow-x:auto;margin:0;\">\n<span style=\"color:#7f848e;\"># Reference by number</span>\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#61afef;\">@claude</span> explain monads\n<span style=\"color:#c678dd;\">gv></span> convert <span style=\"color:#e06c75;\">$1</span> to Python\n\n<span style=\"color:#7f848e;\"># Reference by provider name</span>\n<span style=\"color:#c678dd;\">gv></span> show <span style=\"color:#e06c75;\">$claude</span> <span style=\"color:#7f848e;\"># Last Claude response</span>\n<span style=\"color:#c678dd;\">gv></span> compare <span style=\"color:#e06c75;\">$1</span> and <span style=\"color:#e06c75;\">$2</span>\n</pre>\n</div>\n\n<div style=\"background:#2d2d2d;padding:16px;border-radius:6px;margin-bottom:16px;\">\n<h3 style=\"color:#e5c07b;margin:0 0 12px 0;font-family:sans-serif;\">3. Shell Integration with $refs</h3>\n<p style=\"color:#9ca3af;margin:0 0 12px 0;font-family:sans-serif;\">Use AI responses in Unix commands:</p>\n<pre style=\"background:#252526;padding:12px;border-radius:4px;overflow-x:auto;margin:0;\">\n<span style=\"color:#7f848e;\"># Word count of AI response</span>\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#56b6c2;\">!</span> echo <span style=\"color:#e06c75;\">$1</span> | wc -w\n\n<span style=\"color:#7f848e;\"># Save response to file</span>\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#56b6c2;\">!</span> cat <span style=\"color:#e06c75;\">$claude</span> > output.txt\n\n<span style=\"color:#7f848e;\"># Pipe to system commands</span>\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#e06c75;\">$1.code</span> | python <span style=\"color:#7f848e;\"># Execute code block</span>\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#e06c75;\">$1</span> | pbcopy <span style=\"color:#7f848e;\"># Copy to clipboard (macOS)</span>\n</pre>\n</div>\n\n<div style=\"background:#2d2d2d;padding:16px;border-radius:6px;margin-bottom:16px;\">\n<h3 style=\"color:#e5c07b;margin:0 0 12px 0;font-family:sans-serif;\">4. Compare & Benchmark</h3>\n<p style=\"color:#9ca3af;margin:0 0 12px 0;font-family:sans-serif;\">Compare responses from different providers:</p>\n<pre style=\"background:#252526;padding:12px;border-radius:4px;overflow-x:auto;margin:0;\">\n<span style=\"color:#7f848e;\"># Get responses from both</span>\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#61afef;\">@claude</span> explain monads\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#61afef;\">@openai</span> explain monads\n\n<span style=\"color:#7f848e;\"># Diff them</span>\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#56b6c2;\">/diff</span> <span style=\"color:#e06c75;\">$claude</span> <span style=\"color:#e06c75;\">$openai</span>\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#56b6c2;\">/diff</span> <span style=\"color:#e06c75;\">$1</span> <span style=\"color:#e06c75;\">$2</span> --side-by-side\n\n<span style=\"color:#7f848e;\"># Benchmark all providers</span>\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#56b6c2;\">/benchmark</span> <span style=\"color:#98c379;\">\"explain bubble sort\"</span>\n\n<span style=\"color:#7f848e;\">PROVIDER MODEL LATENCY TOKENS COST</span>\n<span style=\"color:#98c379;\">claude</span> sonnet 1.2s 234 $0.003\n<span style=\"color:#98c379;\">openai</span> gpt-4o 0.9s 198 $0.002\n<span style=\"color:#98c379;\">ollama</span> llama3 2.1s 256 $0.000\n</pre>\n</div>\n\n<div style=\"background:#2d2d2d;padding:16px;border-radius:6px;margin-bottom:16px;\">\n<h3 style=\"color:#e5c07b;margin:0 0 12px 0;font-family:sans-serif;\">5. Export & Import</h3>\n<p style=\"color:#9ca3af;margin:0 0 12px 0;font-family:sans-serif;\">Move data in and out:</p>\n<pre style=\"background:#252526;padding:12px;border-radius:4px;overflow-x:auto;margin:0;\">\n<span style=\"color:#7f848e;\"># Export to file</span>\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#56b6c2;\">/export</span> <span style=\"color:#e06c75;\">$1</span> --format md > explanation.md\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#56b6c2;\">/export</span> session --format json > session.json\n\n<span style=\"color:#7f848e;\"># Import for context</span>\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#56b6c2;\">/import</span> error.log\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#61afef;\">@claude</span> what's causing this error?\n\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#56b6c2;\">/import</span> screenshot.png\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#61afef;\">@claude</span> what's in this image?\n</pre>\n</div>\n\n<div style=\"background:#2d2d2d;padding:16px;border-radius:6px;margin-bottom:16px;\">\n<h3 style=\"color:#e5c07b;margin:0 0 12px 0;font-family:sans-serif;\">6. Remote Multi-Agent</h3>\n<p style=\"color:#9ca3af;margin:0 0 12px 0;font-family:sans-serif;\">Execute across machines:</p>\n<pre style=\"background:#252526;padding:12px;border-radius:4px;overflow-x:auto;margin:0;\">\n<span style=\"color:#7f848e;\"># Run on remote agent</span>\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#56b6c2;\">/remote</span> dev-vm <span style=\"color:#98c379;\">\"@claude explain this codebase\"</span>\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#56b6c2;\">/remote</span> prod-api <span style=\"color:#98c379;\">\"@ollama --model codellama review main.go\"</span>\n\n<span style=\"color:#7f848e;\"># Broadcast to all agents</span>\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#56b6c2;\">/broadcast</span> <span style=\"color:#98c379;\">\"what is your current git branch?\"</span>\n\n<span style=\"color:#7f848e;\">laptop: </span><span style=\"color:#98c379;\">main</span>\n<span style=\"color:#7f848e;\">dev-vm: </span><span style=\"color:#98c379;\">feature/auth</span>\n<span style=\"color:#7f848e;\">prod-api: </span><span style=\"color:#98c379;\">v2.3.1 (tag)</span>\n\n<span style=\"color:#7f848e;\"># Delegate session to another machine</span>\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#56b6c2;\">/delegate</span> dev-vm with <span style=\"color:#98c379;\">\"deploy this fix\"</span>\n</pre>\n</div>\n\n<div style=\"background:#2d2d2d;padding:16px;border-radius:6px;\">\n<h3 style=\"color:#e5c07b;margin:0 0 12px 0;font-family:sans-serif;\">7. Field Extraction</h3>\n<p style=\"color:#9ca3af;margin:0 0 12px 0;font-family:sans-serif;\">Access specific parts of responses:</p>\n<pre style=\"background:#252526;padding:12px;border-radius:4px;overflow-x:auto;margin:0;\">\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#e06c75;\">$1.code</span> <span style=\"color:#7f848e;\"># Extract code blocks</span>\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#e06c75;\">$1.text</span> <span style=\"color:#7f848e;\"># Prose only (no code)</span>\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#e06c75;\">$1.code[0]</span> <span style=\"color:#7f848e;\"># First code block</span>\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#e06c75;\">$1.code.python</span> <span style=\"color:#7f848e;\"># Python code blocks only</span>\n<span style=\"color:#c678dd;\">gv></span> <span style=\"color:#e06c75;\">$1.summary</span> <span style=\"color:#7f848e;\"># AI-generated summary</span>\n</pre>\n</div>\n\n<div style=\"background:#1a3d1a;padding:12px;border-radius:6px;margin-top:16px;font-family:sans-serif;\">\n<h4 style=\"margin:0 0 8px 0;color:#4ade80;\">✨ Key Insight</h4>\n<p style=\"color:#a7f3d0;margin:0;\">gvShell treats AI like Unix: responses are files, providers are commands, and everything composes with pipes. This is the Unix philosophy applied to AI.</p>\n</div>\n\n</div>",
"choices": [
{
"label": "Run gvShell",
"value": "Show me how to run gvShell and try some of these commands",
"primary": true
},
{
"label": "Build status",
"value": "What's the current build status of gvShell - what's implemented?"
},
{
"label": "More examples",
"value": "Show me more advanced gvShell examples from the documentation"
}
]
},
"createdBy": "claude",
"createdAt": "2026-01-06T06:42:02.883Z",
"updatedAt": "2026-01-06T06:42:03.094Z",
"scope": "vibetools",
"tags": [
"gvshell",
"documentation",
"pipes"
],
"targetUser": "claude"
}