Give Claude your brand guidelines one of two ways: paste the written rules into the prompt, or connect a source Claude can read on its own, such as a design brief served over an MCP connection. The first works for a single conversation. The second keeps working after that conversation ends.
Most teams start with the first method without deciding to. Someone pastes a hex value and a line about tone into a chat, Claude writes an on-brand first draft, and the next person on the team starts from nothing, because the brief lived in one conversation nobody else can see. The method you pick decides whether that same work has to happen again tomorrow.
This post covers both: what to write in a brief Claude can use, and how to connect that brief so it reaches Claude before anyone has to paste it.
Key Takeaways
- There are two ways to give Claude your brand guidelines: paste a design brief into the prompt, or connect a source over MCP so Claude reads it before it writes.
- A brief that names a banned word, a tone rule, and a logo spec works for Claude the same way it works for a new hire.
- TrueKeep connects Claude, ChatGPT, and Cursor to a portal through the Model Context Protocol, an open standard, so the brief reaches every conversation, not just one.
- The brand kit skill carries the brief into Claude directly, for people who prompt it all day.
- A brief only helps if someone keeps it current. Claude reads whatever version it finds, including an outdated one.
Two ways to give Claude your brand guidelines
The first way costs nothing to set up. Open a conversation, paste the rules, and ask Claude to write with them. It works immediately, and it stops working the moment the conversation ends. The next person who asks Claude for a headline starts over, and so do you tomorrow.
The second way takes one setup step: connect a source Claude can read on its own. TrueKeep does this through an MCP connection, so Claude reads the brand's strategy, guide, and design brief before it drafts anything, in every conversation, for everyone on the team.
Pasting is right for a one-off: a single email, written once, by someone who will not ask Claude for anything else this quarter. A connection is right for anyone who asks more than once, because the brief only has to be written and kept current in one place.
What belongs in a brief Claude can use
Claude follows instructions literally, so a vague instruction gives it nothing to act on. "Sound professional" could mean anything. A rule like the one below settles the question before Claude writes a word:
Words we avoid: utilize, seamless, world-class.
Use instead: use, clear, trusted.
Formality: direct, no exclamation marks.
Logo: use the stacked mark on light backgrounds only.
Four lines like that decide more drafts than a paragraph about personality. The Mailchimp content style guide is a public example of a brand voice section written at this level of detail, down to how it handles numbers and dates.
This is the same information a style guide gives a new hire. The difference is that Claude cannot ask a colleague what "on-brand" means in practice. Whatever is written down is the whole brief it gets.
Connect Claude to your brand guidelines over MCP
A brief in a document still has to be found, opened, and pasted by a person, every time. An MCP connection removes that step. TrueKeep serves the portal to Claude, ChatGPT, or Cursor through the Model Context Protocol, an open standard for giving AI tools access to a source of information.
Once connected, Claude reads the strategy, the guide, and the design brief before it writes, and a final pass checks the output against the brand and fixes what is off. Nobody has to remember to paste anything, because the guidelines come to the work. The AI connection guide covers the setup steps.
Add the brand kit skill for daily prompting
An MCP connection covers any tool that supports it. For people who prompt Claude all day, every day, TrueKeep also offers the brand kit skill: a downloadable skill that carries the brand into Claude directly.
The skill and the connection solve the same problem for different people. A marketer who opens Claude a dozen times a day wants the skill already running. A team lead setting up Claude for the whole company wants the connection in place once, so nobody has to install anything themselves.
Before and after: what changes in Claude's answer
Ask Claude for a product announcement with no brand context, and it reaches for the words every brand announcement reaches for:
Before: "Supercharge your workflow with our seamless new feature."
After: "The new export adds a CSV option. Turn it on in settings."
The second version is not more careful writing. It is Claude following a brief that names the banned words and shows what to write instead. Nothing about the request changed between the two versions, only whether Claude had the brief before it answered.
The tradeoff: Claude only follows a brief that is current
A connection solves the problem of reaching Claude. It does not solve the problem of being right. Claude reads whatever brief it finds, confidently, whether that brief reflects this month's rules or last year's.
A rebrand, a retired logo variant, or a word the company stopped using can sit in a brief for months before anyone notices Claude still writing with the old one. The fix is the same one that applies to any brand document: name one owner, and give them a reason to check it. Our design brief template covers what that ownership section should say.
FAQ
How do I give Claude my brand guidelines? Paste the written rules into the prompt for a single conversation, or connect a source Claude can read on its own, such as a design brief served over an MCP connection. The second method reaches every conversation without anyone repeating the paste.
Can I just paste my brand guidelines into every Claude conversation? You can, and it works for that conversation. It stops there: the next person, and the next chat, starts from nothing unless they paste the same text again.
What is an MCP connection? MCP, the Model Context Protocol, is an open standard that lets an AI tool read a source of information directly. TrueKeep uses it to connect Claude, ChatGPT, or Cursor to a brand portal, so the AI reads the brand before it writes instead of waiting to be told.
What should a brief for Claude include? The words the brand avoids and the words it uses instead, a line on formality, and any logo or color rule that matters for the request at hand. Specifics give Claude something to follow; adjectives like "modern" do not.
Does Claude remember brand guidelines between conversations? Not on its own. A pasted brief lives in that one conversation. A connected source, through MCP or a skill, is available again the next time, because it reads from the same place rather than from what was typed in.
Does TrueKeep offer a free way to connect Claude? No. TrueKeep is one flat monthly price for the whole organization, with no free plan. See pricing for the current numbers.
Give Claude a brief it can actually use
Claude brand guidelines come down to two choices: paste the rules each time, or connect a source that Claude reads on its own. Write specifics, not adjectives, so there is something for Claude to follow. Decide who owns keeping that brief current, since Claude will read an outdated one with the same confidence as a current one.
If you want the connection instead of the paste, get started with TrueKeep.