One brand voice in three different agents

Marketing and growthMarketing team writing in Claude Code, Codex and Cursor

Our brand rules lived in three files: CLAUDE.md, AGENTS.md and a Cursor rule. They drifted. Cursor still said 'AI-powered' was fine after we banned it. Each file was also loaded in full on every prompt, about 3k tokens.

How do I keep the same brand voice rules in Claude Code, Codex and Cursor without three copies?

Turn your CLAUDE.md, AGENTS.md and Cursor rules into one set of team rules with arcroom import, and accept them on the Rules page. Every agent gets the same rules at the start of each session, in Claude Code, Codex, Cursor or any MCP agent. When a rule changes, edit it on the Rules page. Rules are instructions, not enforcement, so a person still reads before publishing.

Also asked as: Why does Cursor keep using words our brand guide banned? · How do I share one style guide across different AI writing agents? · Can Codex and Cursor read the same brand rules as Claude Code?

The work

Run arcroom import once: each rules file becomes a card on the Rules page, and nothing reaches an agent until a person accepts it. The files are never changed.

ArcRoom keeps
    The next agent gets

    Remember: brand voice: plain words, no 'AI-powered', 'room' not 'workspace', sentence-case headings

    Step by step

    1. Run arcroom import once: each rules file becomes a card on the Rules page, and nothing reaches an agent until a person accepts it. The files are never changed.
    2. Every agent gets the same rules at the start of each session.
    3. Keep one copy in the playbook instead of three files.
    4. When a rule changes, edit it on the Rules page; the old version stays in its history.

    What the agent sees

    Remember: brand voice: plain words, no 'AI-powered', 'room' not 'workspace', sentence-case headings
    Recall (Cursor, landing page prompt): 3 brand rules

    Compared with today's tools

    CLAUDE.md + AGENTS.md + Cursor rules

    Keep the same rules in each tool's own file.

    Setup
    1 hour
    Ongoing
    Edit three files for every rule change

    ArcRoom

    Every agent reads the team playbook before it works and writes to it when it finishes. Nothing to update by hand.

    Setup
    2 min per agent (npx arcroom init --agent …)
    Ongoing
    None

    Rules files cost their full size on every prompt and drift apart; Mem0's OpenMemory shares one store across MCP clients but only when the agent searches, while ArcRoom recall sends the relevant rules every time.

    Where it stops

    Recall is a reminder, not enforcement. A model can still slip, so a human still reads before publishing.

    Try it in your project

    Works with the agents your team already uses, and any agent over MCP or the HTTP API.

    More in marketing and growth