compozy tool invoke
Invoke one registry tool through daemon policy
compozy tool invoke
Invoke one registry tool through daemon policy
compozy tool invoke <tool_id> [flags]Examples
# Invoke a tool with inline JSON input
compozy tool invoke compozy__tool_info --input '{"tool_id":"compozy__skill_view"}' -o json
# Invoke a tool with JSON read from a file
compozy tool invoke compozy__tool_info --input-file ./input.json -o json
# Invoke a tool with JSON read from stdin
echo '{"tool_id":"compozy__skill_view"}' | compozy tool invoke compozy__tool_info -o jsonOptions
--agent string Agent name for scoped diagnostics
--approval-token string Single-use approval token for approval-gated tools
--correlation-id string Optional correlation id
-h, --help help for invoke
--input string Inline JSON input
--input-file string Path to JSON input file, or '-' for stdin
--sensitive-input-field stringArray Input field path to redact in events
--session string Session id for scoped diagnostics
--tool-call-id string Optional caller tool-call id
--turn-id string Optional caller turn id
--workspace string Override workspace diagnostics (ID, name, or path)Options inherited from parent commands
--json Emit JSON output
-o, --output string Output format: human, json, jsonl, or toon (default "human")Output Formats
The root -o, --output flag accepts these values. A renderer is supported only when this command writes that result; jsonl is for commands that emit one JSON record per line.
| Format | Command result |
|---|---|
human | Interactive terminal rendering |
json | Machine-readable JSON rendering |
jsonl | One JSON record per line when supported |
toon | Compact agent-readable rendering |
Example:
compozy tool invoke <tool_id> -o json