lockstep
4 min readNaman Jain

Claude Code output styles: built-in styles and custom examples

What Claude Code output styles do, the built-in Proactive, Concise, Explanatory, and Learning styles, how to switch, and how to write a custom style.

Some people want Claude Code to stop narrating and just give the result. Others, learning a new codebase, want every change explained. Repeating "be brief" or "explain why" in every prompt is tedious. Output styles set it once for the session.

This guide covers the built-in styles, switching, and custom styles, from Anthropic's output styles documentation, checked on 16 November 2026. Output styles are a current, documented feature; the "deprecated" you may see in search suggestions doesn't match Anthropic's docs as of that date.

The built-in styles

Claude Code starts in Default, its standard instructions for software engineering. Four built-in styles keep those instructions and add their own:

StyleWhat changesUse it when
ProactiveClaude starts work right away and makes reasonable assumptions instead of asking about routine decisionsYou'd rather correct course than answer questions
ConciseResponses lead with the result, without preamble, narration, or recapDefault responses are longer than you want
ExplanatoryShort Insight blocks explain the choices behind the codeYou're learning a codebase
LearningInsight blocks, plus Claude leaves small pieces of code for you to writeYou want hands-on practice while the task gets done

A few details matter:

  • Concise still writes in full when you ask for detail, and for anything you need to act safely: errors, failing test output, security warnings, and confirmations before destructive actions.
  • Proactive doesn't change your permission mode. Prompts appear as before. Its instructions also tell Claude to check with you before deleting data or changing shared or production systems.
  • Learning marks the spots it leaves for you with a TODO(human) comment, then waits until you say you're done.

Switching styles

  • /output-style concise switches; /output-style alone lists styles. The command ignores case.
  • /config, then Output style, picks from a menu.
  • In a settings file: "outputStyle": "Concise". The value is case-sensitive, and a name that doesn't match exactly gives you Default.

The command and menu save to .claude/settings.local.json. To make a style your default everywhere, set outputStyle in ~/.claude/settings.json; a project's own settings take precedence. A switch takes effect from your next message.

Writing a custom style

A custom style is a markdown file:

  • ~/.claude/output-styles/ for you, in every project
  • .claude/output-styles/ for a project
  • A managed settings directory for an organization

The file name is the style name unless you set name. An example that keeps Claude coding normally but changes how it reports:

---
name: Reviewer notes
description: End every change with a short review checklist
keep-coding-instructions: true
---

After each change, end with a section titled "Review" listing:
- the files changed and why, one line each
- what to test by hand
- anything that might affect other teams
Keep the rest of the response short.

keep-coding-instructions matters. Custom styles drop Claude Code's built-in software engineering instructions, such as how to scope changes and verify work, unless this is set to true. Leave it out only when Claude won't be writing code, for example a style that turns it into a writing assistant.

Style files are read at startup, so restart Claude Code after creating or editing one. Plugins can ship styles too, in an output-styles/ directory.

Output style or something else?

From Anthropic's comparison:

You wantUse
Every response in a certain voice, length, or formatAn output style
Claude to know your project's conventions and commandsCLAUDE.md
Instructions for one kind of taskA skill
Something to happen every time, without exceptionA hook
A helper with its own tools and modelA subagent

Output styles apply to the main conversation, not to most subagents, which run their own system prompt. Explanatory and Learning produce longer responses and more output tokens; Concise produces fewer.

Where Lockstep fits

Output styles change how Claude talks. Lockstep changes what it knows about your team's decisions: it captures them from PRDs, Slack, and pull requests, makes each binding when its owner approves it, and briefs Claude Code on the ones that apply at the start of each session, whichever style you use.

See how it works.

Sources checked 16 November 2026

For choosing between every extension, see CLAUDE.md vs skills vs hooks vs subagents. For settings files, see Claude Code settings.json.

Frequently asked questions

What are Claude Code output styles?
Instructions that set Claude's role, tone, and response format for every response in a session, so you don't repeat the request in each prompt. Claude Code has four built-in styles besides Default, and you can write your own.
Are Claude Code output styles deprecated?
No. As of 16 November 2026, Anthropic's documentation describes output styles as a current feature, with recent additions such as the Concise style and the /output-style command working in non-interactive mode.
How do I change the output style in Claude Code?
Run /output-style followed by a style name, such as /output-style concise, or pick one in /config. You can also set outputStyle in a settings file; the value is case-sensitive, so write Concise, not concise.
How do I make a custom output style?
Create a markdown file in ~/.claude/output-styles/ for yourself or .claude/output-styles/ for a project. Add frontmatter with a name and description, set keep-coding-instructions: true if Claude should still code normally, and write the instructions below. Restart Claude Code to pick up a new file.
Is an output style the same as CLAUDE.md?
No. CLAUDE.md holds what Claude should know about your project. An output style changes how Claude responds. Both are instructions Claude follows, not guarantees; use a hook for anything that must always happen.

Keep reading