claude-exec-json

Run Claude Code in exec mode and write structured JSON output to a temp file.

Install / npx

npx -y claude-exec-json "summarize this repo"

Usage

claude-exec-json [options] <prompt...>

Options:

• --schema <json> Inline JSON schema string
• --schema-file <path> Path to JSON schema file
• --out <path> Output file path (default: /tmp/.claude-exec.<uuid>.json)
• --model <name> Claude model alias or full name
• --max-turns <n> Maximum number of agentic turns
• --claude-bin <path> Path to claude binary (default: claude)
• --yolo Use --dangerously-skip-permissions
• --help Show help

Pass through args to claude after --:

claude-exec-json -- --allowedTools Bash,Edit "update the README"

For LLMs / agents

This tool is designed for agents that need a one-shot, non-persistent run and a JSON file on disk. It prints the output path on stdout, so you can read the file next.

npx -y claude-exec-json --schema '{\"type\":\"object\",\"properties\":{\"summary\":{\"type\":\"string\"},\"next\":{\"type\":\"array\",\"items\":{\"type\":\"string\"}}},\"required\":[\"summary\",\"next\"]}' \"Summarize and list next steps\"

Example flow:

OUT=$(npx -y claude-exec-json \"Describe the repo in JSON\")
cat \"$OUT\"

If you use the Codex skills format, a ready-to-pull skill is included at skills/claude-exec-json/SKILL.md.

Default schema

If no schema is provided, this default is used:

{
  "type": "object",
  "additionalProperties": false,
  "properties": {
    "ok": { "type": "boolean" },
    "result": { "type": "string" },
    "error": { "type": "string" }
  },
  "required": ["ok", "result"]
}

Example with schema file

cat > /tmp/schema.json <<'JSON'
{
  "type": "object",
  "additionalProperties": false,
  "properties": {
    "summary": { "type": "string" },
    "tasks": { "type": "array", "items": { "type": "string" } }
  },
  "required": ["summary", "tasks"]
}
JSON

claude-exec-json --schema-file /tmp/schema.json "List next actions"

The command prints the output path on stdout.