Text version of the 17 slides. Status labels: shipped means verified in the current product; in development means code exists but is not released or proven; planned means designed but not built; hypothesis means an assumption not yet tested; historical means true once and no longer current.
Inkfold
Inkfold
Every major model in one workspace, with a memory that follows you. Built on the FusionLayer engine.
The one-minute version
Inkfold puts ChatGPT, Claude, Gemini and other models in one workspace, with one memory they all share.
It gives each model the right context, can put several models to work on one question, and lets each person choose what it keeps.
The need
Several assistants, no shared memory
People now use more than one assistant, and each one starts from nothing.
Starting over in each assistant
A trip, a thesis or a client project is explained to ChatGPT, then again to Claude, then again to Gemini.
Memory locked in one app
Each vendor's memory stays inside its own app. Other models never see what was settled there.
One model's view
For a question that matters, a single model gives a single opinion, with no second view to check it against.
Who it serves
People who use more than one assistant
Consumers and prosumers first; small teams can share a workspace and its memory.
Writers and researchers
Long projects with sources, notes and arguments carried from one session to the next, months apart.
Students
Courses, reading lists and drafts remembered across whichever model helps with each task.
Freelancers and independent professionals
Each client's tone, constraints and history kept separate, so one client's context never reaches another's.
Small businesses
Pricing, policies and standard answers settled once, and reused by everyone who asks a model for help.
How it works
From a conversation to a considered answer
- Every model in one place ChatGPT, Claude, Gemini and more, with own keys or one OpenRouter key.
- Memory kept Projects, preferences and settled facts, each with its source.
- Context added The relevant memories reach each new question automatically.
- Several models Compare, Council, Debate or Pipeline when a question matters.
- A privacy choice Smart, Private or Incognito, per conversation.
Capability one
A memory that carries across models
ChatGPT today and Claude tomorrow both start from what was already said.
shipped What matters, kept
Projects, preferences and the facts a person has settled, each with its source.
shipped A map of the context
People, places and topics linked, so related facts are found together.
shipped Receipts and correction
Each answer shows the memories it used; one click marks one wrong, outdated or only for one project.
shipped Past chats imported
History from ChatGPT, Claude, Gemini, Copilot and Grok exports, so nobody starts from zero.
Status labels: shipped means verified in the current product; in development means code exists but is not released or proven; planned means designed but not built; hypothesis means an assumption not yet tested; historical means true once and no longer current.
Capability one, continued
The right context for each question
Inkfold chooses which memories a question needs, so the model gets a focused brief.
shipped Chosen automatically
Inkfold reads the question, searches the context map and adds only what is relevant. Nothing is picked by hand.
shipped Within a size budget
The model receives a short, focused brief rather than a person's entire history.
shipped Kept apart by project
A question about one client adds that client's facts only; another client's project never enters the context.
Capability two
The right model for each question
Every major vendor in one place, with the cost of each model in view.
shipped Every major vendor
OpenAI, Anthropic, Google, DeepSeek, Mistral, xAI and more, in one workspace.
shipped Own keys, or one key
A person's own key for each vendor, or a single OpenRouter key that reaches all of them.
shipped Spending view
What each model has cost, with budget rules that cap a month before it surprises anyone.
planned Automatic model choice
Inkfold picking the model for each question by task and cost.
Capability three
Several models, one considered answer
For questions that matter, Inkfold asks more than one model.
| Mode | What it does | Everyday example |
|---|---|---|
| Compare | Answers from several models, side by side | A researcher reads two summaries before deciding which citations to trust |
| Council | Several answers gathered into one | A small business merges several drafts of a refund policy into one version |
| Debate | Models challenge each other until they agree | A freelancer checks whether a contract clause is standard |
| Pipeline | Stages: plan, draft, check and polish | A trip itinerary is planned, drafted, checked for gaps and polished |
All four modes are shipped in the beta. Examples from the How it works page, inkfold.app, 28 September 2026.
Capability four
Better prompts and ready-made ways of working
A rushed question becomes a clear one, and good ways of working are shared.
shipped Prompt improvement
The composer improves, rephrases or tunes a prompt for the chosen model, before it is sent.
shipped Shared library, in beta
Personas, playbooks, routines, templates, guidelines and packs that people publish, rate, save and adapt.
planned Paid library items
Items that creators sell, with revenue sharing for the people who publish them.
Capability five
A privacy choice for each conversation
Three modes, chosen for each conversation and free on every plan.
| Mode | What is kept | What Inkfold can read |
|---|---|---|
| Smart (default) | Stored and indexed, encrypted at rest, to build memory | The content, so it can find the right context |
| Private | Encrypted in the browser, with a key made from a phrase only the person knows, before it is stored | Nothing: only ciphertext; memory features are limited |
| Incognito | Nothing after the conversation ends | Nothing is kept |
In every mode, the text of a request goes to the model vendor the person chose. Model keys are stored encrypted; memory can be exported or deleted at any time.
Alternatives
Alternatives people use today
| Option | What it gives | What it leaves out |
|---|---|---|
| One vendor's app: ChatGPT, Claude or Gemini | Memory inside that one app | Other models never see it |
| Many-model chat apps: Poe, TypingMind, OpenRouter Chat | Many models behind one screen | No shared memory; no models working together |
| Developer memory services: Mem0, Supermemory | A memory store other software calls | Built for developers, not people |
| Inkfold | One memory across models, several models on one question, privacy per conversation | Automatic model choice is coming |
Other products described from their own published material, September 2026; not tested side by side.
What makes it different
Four advantages that work together
How Inkfold sees its difference.
One memory, every model
What was settled with one vendor's model is known to the next, with a source and a receipt.
Models working together
Compare, Council, Debate and Pipeline across vendors, not just a menu of models.
Privacy never paywalled
Smart, Private and Incognito for each conversation, free on every plan.
An engine underneath
Built on FusionLayer, the orchestration and context engine that also serves organisations and developers.
Ways to use Inkfold
Web today, more surfaces coming
shipped Web application
In any desktop browser at app.inkfold.app, with no download.
in development On a phone
Installable from the phone's browser to the home screen.
in development Windows desktop
Local files to ask about, tray controls and an identity that needs no email. Invited testers today.
in development Team workspaces
A workspace and memory a small team shares, now in beta.
Business model
Planned pricing
Subscriptions; privacy modes and own keys are free on every plan.
| Plan | Planned price | What it adds |
|---|---|---|
| Free | $0 | One memory across models, receipts and correction, the shared library, all three privacy modes |
| Pro | $19 a month or $190 a year | Multi-model modes, prompt improvement, spending view with budget rules, own storage |
| Team | $32 per seat a month | Team workspaces and shared memory, admin console and activity log |
| Founding member | $190 a year, first 20 people | Pro, with the price locked while subscribed and a direct line to the founder |
Planned prices as shown on inkfold.app, 28 September 2026. Nothing is charged yet; checkout opens with the founding cohort.
Roadmap
What comes next, in order
shipped Today
- Web application in private beta
- Memory, four multi-model modes, library, privacy modes
in development Next
- Installable phone web application
- Signed Windows desktop app
planned Then
- Automatic model choice by task and cost
- Team workspaces out of beta
planned Later
- Paid library items with revenue sharing
- Inkfold-managed model access
Inkfold
Every major model in one workspace, with a memory that follows you.
Private beta, by invitation. Built on the FusionLayer engine. inkfold.app
Facts as of 28 September 2026. Status labels: shipped, in development, planned.