Portability & ownership

Your context should outlive your model.

Mara does not host or run AI models for you. It writes readable files you can take to whichever AI you already use — and keep when you change your mind.

Open-weight ready

Your context should outlive your model.

A Mara Playbook is made from readable, portable files—not hidden memory inside one platform. Use it with a hosted assistant today, an open-weight model tomorrow, or infrastructure you control.

Running a model on infrastructure you control can keep your context inside systems you administer. That depends on how the deployment is configured — an open-weight model is not automatically private or secure.

Portable

Move your Playbook when you change models.

Inspectable

Read and edit every instruction and fact.

Scoped

Share only the context needed for the current job.

Owned

Keep the resulting files independently of the AI platform.

Model intelligence is becoming abundant. The context that makes it useful is the part worth owning.

What actually gets moved

  • Hosted assistants. Paste the copy-ready block into ChatGPT or Gemini, or drop the folder next to a Claude project. No plugin, no integration.
  • Private or company deployments. The same plain text can be used inside an environment your organisation controls, subject to its own policies.
  • Local open-weight models. Because the Playbook is text, a model running on your own hardware can read it exactly as a hosted one would.

Running a model on infrastructure you control can keep your context inside systems you administer. That depends entirely on how the deployment is configured — an open-weight model is not automatically private or secure.

What Mara keeps

Your confirmed context and completed Playbooks stay in your account so you can edit and download them. Raw interview messages exist only long enough to compile your Playbook and are cleared once it replaces them. You can delete a Playbook or your whole account at any time. See privacy & terms for the specifics.