Independent multi-AI context layer
Write your context once. Use it everywhere.
Configure ChatGPT, Claude and Gemini around the way you work, and keep them up to date when the platforms change.
The diagnostic is not open yet. This page describes exactly what it will produce.
A register kept properly: an entry, its date, its source
3 platforms in V1
ChatGPT, Claude, Gemini
1 source of truth
a versioned profile that belongs to you
0 mega-prompt
each piece goes where it belongs
Still repeating the same information to ChatGPT, Claude and Gemini?
Who you are, who you serve, the form the output must take, what must never be said, the client nobody names. That context lives in your head and in dozens of pasted paragraphs.
01The mega-prompt
One giant text pasted everywhere. It exceeds the limits, mixes durable facts with temporary ones, and ages the moment an interface changes.
02Native memory
Each platform accumulates its own memory, which you do not choose, cannot audit, and which stays with the vendor.
03Prompt libraries
They sell more text, not a structure. They never say what is durable, where to paste it, or what fits within the limit.
The valuable asset is not the prompt. It is the structured context profile — and compiling it for each platform must be deterministic.
One diagnostic, one profile, one compilation per platform
10 to 12 questions with clickable examples. No account and no payment before the result. Target: 8 minutes from the first question to an installed configuration.
- 01DiagnosticRole, objectives, audience, recurring tasks, formats, tone, rules, mistakes to avoid, projects. One question per screen, never a large empty box.
- 02Privacy reviewNames, emails, phone numbers, clients and identifiers are detected and masked before any call to a model. You unmask, or you mask more.
- 03Context ProfileA structured, versioned object: every field carries its scope and its stability. You correct it line by line.
- 04CompilationDeterministic. Classification, destination and limits are never decided by a model. Models extract and rephrase, nothing more.
- 05Installation guidePer Configuration: what, why, where exactly inside the platform, and one line to tick. Then three witness prompts to verify.
Compilation
ChatGPT
- Global Instructions
- Project Instructions
- Memory Recommendation
Claude
- Global Instructions
- Project Instructions
- Style
- Skill
Gemini
- Memory Recommendation
- Gem
- Gem Instructions
ChatGPT
- Global Instructions
- Project Instructions
- Memory Recommendation
Claude
- Global Instructions
- Project Instructions
- Style
- Skill
Gemini
- Memory Recommendation
- Gem
- Gem Instructions
Capabilities and destinations to be verified for your country and your plan. Every registry entry carries its source and its date.
What you receive, platform by platform
One card per Configuration, never one blob. Each card carries its destination, its reason, its character count against the limit, its version and its date.
Where: account settings, custom instructions.
1,842 / 3,000 characters — fits within the limit
v3 · verified 21 August 2026
Where: account settings. A SKILL.md file to upload.
to be verified for your country and your plan
v3 · verified 21 August 2026
Example. Real limits and installation paths are re-read at every compilation, with their source and their date.
The promise is kept when you tick the last line of the guide.
The page confirms “Configuration installed”. Your working rules now live inside the platform, in the right places, without a single pasted mega-prompt.
What leaves, and what never will
One screen before any call to a model, never skippable, never set to “send everything”. We do not keep your raw text.
What will be sent
Scoping memo for [CLIENT_1], sent to [CONTACT_1] ahead of the board meeting.
2 masked spans · pseudonymised text
What will never be sent
- excludedYour client's name
- excludedAddresses and phone numbers
- excludedIdentifiers and secrets
- excludedAny field marked do_not_remember
The platforms change. Your context does not.
We keep a register of each platform's capabilities: limits, installation paths, availability by country and plan, source and verification date. When an entry moves, we know which profiles are affected.
Version history
Nothing is overwritten. Restoring an older version creates a new version.
Platform updates
Installation path for one capability changed. 2 of your Configurations are affected.
Character limit for custom instructions to be re-verified.
Availability of one capability extended to new countries.
Recompiling what moved, and nothing else: “3 items to reinstall, 5 unchanged”.
The only promise we keep by email: we write to you when one of your platforms changes. Not before.
Passes, not a subscription
One-time purchase, duration stated next to the price. At expiry the account returns to Free and the profile is never deleted.
Free
0 €
no time limit
- 1 context profile
- Privacy review
- 1 platform of your choice, compiled and installable
- Installation guide and witness prompts
Pass Pro Founder
39 €
one time, 12 months of access
- All 3 V1 platforms
- Several profiles and versioning
- Recompilation and guides kept current
- Platform update alerts
reserved for the first 100 buyers
Pass Team Founder
290 €
one time, 12 months, up to 5 members
- Workspace
- Company Context and Role Pack
- Shared rules and administration
- Member onboarding
Pass Pro at 69 €, one time, 12 months, is prepared and will be activated after the first 100 passes. Purchase opens with the product.
Teams
Four layers, kept separate by the engine. This is the layer no vendor's personal memory will provide.
Clarifications
What we claim, and what we do not. The position is narrow, and true.
This is not a prompt library
We do not sell text. We keep a structure; the text is its output, disposable and recompilable.
Models decide nothing
They extract and rephrase, on masked text. Architecture, scope, destination and limits are deterministic.
No API to wire up
Everything happens in the chat window, with exact paths and a copy button.
Copy-paste stays manual
We have no access to your ChatGPT, Claude or Gemini account, and pretending otherwise would be a lie.
No measured gain promised
We state what the product does and what it does not. Anything uncertain is marked uncertain.
You leave with your profile
Export of the context profile and of every Configuration. Permanent deletion, stated plainly before it happens.
Next step
Your AI can remember. McKyntext decides what deserves to be.
The diagnostic opens soon: 10 to 12 questions, a privacy review, a versioned profile, one platform configured.