An AI knowledge base your marketing team owns

Every marketer has done this. Open ChatGPT, paste the brand guide, paste the positioning doc, paste last quarter's launch email "for tone". Ask for a draft. Get something close but not quite you.

Then a teammate does the same thing in Claude, with an older version of the guide.

The AI isn't the problem. It just doesn't know your company. Gryz gives your team one shared place it can learn from, in every tool.

Where brand knowledge lives today

• In people's heads. "We never say 'revolutionary'" is a rule everyone knows except the new hire.
• In scattered docs. Three versions of the messaging framework, one of them current.
• In old chats. The best launch email prompt lives in someone's ChatGPT history.
• Nowhere at all. Why you dropped a tagline, or which campaign flopped and why.

What goes into Gryz

In a company workspace, a marketing team might set up a Brand project (open to everyone) and a Campaigns project. Then:

• Documents. Brand guidelines, voice and tone, positioning, the messaging framework, briefs and past campaigns, each with version history.
• A North Star on the project. A short note your people write, up to 1,500 characters: who you're for, what matters most, what never to do. Agents with access to the project read it first.
• Memory. Decisions as they happen. "Launch emails go out on Tuesdays." "We stopped using the word 'seamless'." Each one shows who saved it.
• A company skill. Owners and admins can publish a brand-voice rule that agents read before they write. Nothing reaches an agent until a named person publishes it.
• Tasks. "Draft the launch email", with an owner, a due date and a reviewer.

If you already keep some of this in your own Gryz, Bring your work in moves it on purpose, after showing you exactly what will change.

How your AI tools use it

Once a company owner turns on connected apps, each person connects their AI tool to Gryz once and adds the company. Under the Team policy, Claude, ChatGPT and Cursor are allowed straight away.

• Reading. The tool searches the company's documents, memory and tasks, and opens the right passage. Gryz's search and fetch use the names ChatGPT's deep research and company knowledge expect, so they work from ChatGPT too.
• Writing back. A company starts read only. If the company allows writing and you tick Allow writing for that app, it can create and edit company documents in projects you can edit.
• Suggesting instead. The person who wrote a document, or an owner or admin, picks what assistants may do with it: edit directly, only suggest changes, or read only. A company can also say apps may only suggest. Suggestions wait for a person to accept.

Nobody gets more than they could already open. A restricted project stays inside its people, for agents too.

Example: a launch email

Here's how it can look. (A made-up example.)

1. A writer asks ChatGPT: "Draft the launch email for the new reporting feature, in our voice."
2. ChatGPT searches the company space. It finds the voice guide, the positioning doc and last spring's launch email, and reads the brand-voice skill.
3. It drafts the email in a new company document in the Campaigns project.
4. The brand lead asks Claude to tighten the subject line. Claude sends it as a suggested change, and she accepts it on the document's edit page.
5. The writer asks ChatGPT to remember: "Feature launches lead with the customer problem, not the feature name."

Next time anyone drafts a launch email, in any tool, that decision is already there.

Not just marketing

The same setup works for any team whose work depends on shared knowledge:

• Sales. Positioning, objection handling and competitor notes in one project, so every rep's assistant answers the same way.
• Ops. Procedures as documents, recurring work as tasks with a named reviewer, decisions as memory.
• Product. Specs, decisions and a project North Star, read by product and engineering in different tools. Developers, see Shared context for AI coding agents.

The company stays in control

• Off until an owner says so. Connected apps start off for every company.
• Two policies. Team for simple setup, Enterprise when every app needs approval first. See Team and Enterprise.
• Keep sensitive work out. Mark a project No connected apps.
• No public links. While an app can see the company, it can't publish anything publicly.
• A record. Activity shows which app or agent used which tool, in which project, and when, without the text.
• Private stays private. Each person's Private space is theirs. The company sees only what goes into the company. See What is private in a company workspace.

Honest limits

• Your AI provider sees what it reads. When ChatGPT or Claude reads company text, it goes to that provider under your own agreement with them. Gryz can't control what the app does with it afterward. Read AI apps and your company before you turn things on.
• It's for knowledge, not assets. Gryz keeps text documents like Markdown and HTML. It isn't a place for your logo files and photo library, and images aren't searchable.
• Pictures from other websites don't load in company documents. Embed them in the document instead.
• Company search is off until an owner or admin turns it on. It sends short passages of text to Cloudflare, our AI provider. See How search works.
• Gryz is hosted in the United States.
• Rolling out. Company workspaces are rolling out, so some of this may not be on for you yet.

Get started

• Try it on your own work first. Connect Gryz to ChatGPT or Claude and import a few of your docs. Chat with your documents from any AI tool shows how.
• Set it up for your team. Email info@grtwo.com and we'll create your company workspace with you.

Why this matters beyond one team is in Context compounds.