Use --output-format json to Get Structured Output for Scripts and Automation

Claude Code's default output is human-friendly text, but when you're piping results into another script or building an automated pipeline, you need something a machine can parse. That's what --output-format json is for.

# Get structured JSON output from a one-shot command
claude -p --output-format json "list all TODO comments in the codebase"

The output includes structured data about every step Claude took — the messages, tool calls, file reads, and the final response — all as parseable JSON. Your scripts can extract exactly the fields they need.

This pairs naturally with -p for automation workflows:

# Extract just Claude's answer using jq
claude -p --output-format json "what's the primary database driver?" \
  | jq -r '.[-1].content'

# Build a script that processes Claude's structured output
claude -p --output-format json "find all deprecated functions" \
  | jq '[.[] | select(.type == "assistant")] | last'

You can also use --output-format stream-json for streaming JSON — each line is a separate JSON object emitted as Claude works, so your script can process results in real time instead of waiting for the full response:

# Stream JSON events as they happen
claude -p --output-format stream-json "refactor the auth module" \
  | while read -r line; do
    echo "$line" | jq -r '.type'
  done

Practical automation use cases:

• CI pipelines — parse Claude's code review into structured findings
• Custom dashboards — extract metrics from Claude's analysis
• Chained tools — feed Claude's JSON output into other CLI tools
• Logging — capture structured records of what Claude did and why

Plain text is for humans, JSON is for machines — use --output-format json when your script is the one reading.

via Claude Code