CLI output

Human-readable output

The default output is short and designed for a terminal:

npx @genie-react/cli call react_get_renders '{"limit":5}'

One JSON value

Use --json when the next step needs the full result:

npx @genie-react/cli call react_get_renders '{}' --json | jq '.coverage'

Selected fields as JSONL

Use --fields to keep only the records the agent needs:

npx @genie-react/cli call query_list '{}' \
  --fields queryHash,status,fetchStatus

{
  "queryHash": "[\"demo\",\"greeting\"]",
  "status": "success",
  "fetchStatus": "idle"
}

Batch calls

Batch calls share one connection and run in order:

npx @genie-react/cli batch \
  '[{"tool":"react_get_tree","args":{"depth":2}},{"tool":"react_get_renders","args":{"limit":5}}]' \
  --ndjson

• batch and batch --ndjson write one JSON object per line.
• batch --json writes one JSON array.
• Batch continues after an item fails and exits non-zero if any item failed.
• CLI status, batch, and error envelopes include schemaVersion.

Diagnose startup

npx @genie-react/cli status --verbose

Verbose startup details go to stderr. JSON stdout stays valid.

How this helps the agent

The agent can request only the data it needs, pipe it through normal shell tools, and recover from a stable machine-readable error without parsing prose.