Context layer for your content
Everything we write for you runs on one set of rules: a superprompt with your positioning, voice and the facts that have to be right, editorial guidelines for each format, and the checks a piece passes before it goes live. On request your content team can work from the same layer in Claude or ChatGPT, so in-house writers, freelancers and other agencies all write to one standard.
Why content needs one
More people and more AI tools write for you than ever. Most of them start from a different brief.
Every writer starts from a different brief
The in-house team, two freelancers and an agency each got a different version of how you sound, and every AI chat starts from nothing until someone pastes something in.
Guidelines sit in a PDF
The brand book exists, but nobody opens forty pages before writing a comparison page. So the guidance is there and the content ignores it.
Facts drift from piece to piece
A price changes, a product gets renamed, legal corrects a claim. Old versions keep turning up in new content, and in what AI engines then say about you.
What is in the context layer
Built with you, kept current as we work, and scoped to one job: making every piece of content right before it ships.
Superprompt
One source of truth for how you write about yourself
The superprompt is what every piece reads first: your company, your audience and positioning, your voice, and the facts that have to be right every time. It also lists the sources you trust, so research starts from pages you approved rather than whatever ranks.
What it holds one document per company
- Company
- What you sell, to whom, and how you are different
- Audience
- Your buyers and the words they use
- Voice
- How you sound, and the words you never use
- Facts
- Prices, product names, numbers and approved claims
- Sources
- The pages and publications your brand treats as authoritative
Editorial guidelines
How each kind of piece gets written
A format says what a guide, a comparison or a case study looks like for you. A playbook says how a kind of work runs, from a new article to a YouTube description. Both start from our defaults and take your own rules on top, so a comparison page reads like yours and not like everyone else's.
Formats what the result looks like
- Guide
- Listicle
- Comparison
- Case study
Playbooks how a kind of work runs
- New content
- Content updates
- Programmatic pages
- YouTube
Pre-publishing checks
What a piece has to pass before it goes live
Guidelines only help if someone checks against them. Before a draft reaches you, it goes through the same checks every time: facts against the superprompt and against what your other pages already say, sources for every number, voice, your brand rules, and the technical basics. When a check finds a number that disagrees with another page, we fix every page that has it, not only the new one.
Checked every time by us, before you see a draft
- Facts against the superprompt and against your other pages
- Every claim with a number has a source
- Voice, and the phrases that make text read as written by AI
- Your brand rules, including what you never say
- Structure, SEO and AI Search basics
Access
The same rules in your team's own assistant
The layer lives in the Radyant Cockpit, the operating system we built for our own team, and the Cockpit has an MCP server, the standard way for an AI assistant to reach outside tools. On request we give your content team access, so a writer in Claude or ChatGPT drafts from the same facts, voice and guidelines we do, instead of pasting a brand book into every chat.
Connected over MCP on request
What your writers can do from inside the chat
- Load the superprompt before drafting
- Work to the same format and playbook we use
- Check a claim against your approved facts
- GDPR compliantWe handle your data carefully and provide a DPA on request
- Hosted in GermanyWe use German Hetzner servers in Falkenstein
- EU-hosted AI modelsFrontier models from Claude, Gemini & Mistral
Who writes to it
Your in-house content team
Drafts in their own assistant with your voice and facts already loaded, instead of rebuilding the brief every time.
Freelancers and other agencies
Get the same guidelines and the same checks, instead of one version per person who happened to onboard them.
Us
Every guide, comparison, case study and update we write for you starts from the layer and goes through its checks.
Questions about the context layer
What it covers, what it does not, and how your team gets access.
Is this a wiki or a knowledge base for our whole company?
No. It covers one job: the content we and your writers produce. It holds what a writer needs to write for you and what a piece has to pass before it goes live. Your company documentation, your wiki and your project tools stay where they are.
How is this different from a brand book?
A brand book describes how you want to sound. The context layer is written to be used while writing, by a person or an AI assistant, and it holds the facts and the checks as well as the voice. It is kept current as we work instead of being rewritten once a year.
How does our team get access?
On request. The layer lives in the Radyant Cockpit, and the Cockpit has an MCP server that assistants like Claude and ChatGPT can connect to. When you ask for it, we set up access for your content team and walk them through connecting it.
Who keeps it up to date?
We do, as part of the work. When legal corrects a claim, a product gets renamed or a customer call shows how buyers actually describe you, it goes into the layer, and the next piece anyone writes starts from the new version.
Where is our data stored?
On our own servers at Hetzner in Germany. We handle your data under GDPR and provide a data processing agreement on request.
