Running Capabilities¶

Once you have a Manifest, intent-os run executes it on any available runtime.

Basic Usage¶

# Run from a file
intent-os run examples/translate.yaml --input '{"text": "Hello", "target_lang": "zh"}'

# Run a built-in capability by name
intent-os run translate --input '{"text": "Hello", "target_lang": "zh"}'

# Run with inline parameters
intent-os run translate -p text="Hello" -p target_lang=zh

# Run with inline text (maps to the "text" field)
intent-os run translate "Hello" -p target_lang=zh

Selecting a Runtime¶

# Auto-select (uses available adapters in order)
intent-os run translate -p text="Hello" -p target_lang=zh

# Specify a specific adapter
intent-os run translate --adapter openai -p text="Hello" -p target_lang=zh
intent-os run translate --adapter ollama -p text="Hello" -p target_lang=zh
intent-os run translate --adapter anthropic -p text="Hello" -p target_lang=zh

Input Options¶

# JSON string
--input '{"text": "Hello", "target_lang": "zh"}'

# JSON file
--input-file input.json

# Key=value pairs (repeatable)
-p text="Hello" -p target_lang=zh

# Inline text (positional argument — maps to "text" field)
intent-os run translate "Hello world" -p target_lang=zh

Output¶

# Print to stdout (default)
intent-os run translate -p text="Hello" -p target_lang=zh

# Save execution record to file
intent-os run translate -p text="Hello" -p target_lang=zh --output result.json

# Save execution record
intent-os run translate -p text="Hello" -p target_lang=zh --save records/

Available Adapters¶

Adapter	Provider	Requires
`ollama`	Ollama (local)	`ollama serve` running
`openai`	OpenAI API	`OPENAI_API_KEY` env var
`anthropic`	Anthropic API	`ANTHROPIC_API_KEY` env var
`openrouter`	OpenRouter API	`OPENROUTER_API_KEY` env var
`github-models`	GitHub Models	`GITHUB_TOKEN` env var
`simulated`	Built-in (no AI)	None (auto-fallback)

Compare Across Runtimes¶

intent-os compare examples/translate.yaml --input '{"text": "Hello", "target_lang": "fr"}'

This runs the same capability on every available adapter and compares latency, cost, and token usage.