MokenLogic

Memory API / Developers

Planned · No public availability date

Give any agent the context its user already built.

The planned MokenLogic Memory API will let an authorized product keep a separate memory for each of its users, and retrieve relevant, sourced context for whichever model it runs.

StatePlanned
ReturnsSourced memories
ModelsAny

How it will work / 01

Send sessions. Ask for context.

01

Send

Your product sends finished sessions and documents for each user.

02

Update

That user’s memory updates regularly in the background.

03

Query

Before a model call, request context for the task at hand.

04

Use

Pass the returned memories, with their sources, to any model.

What you get back / 02

Memories with sources, not a generated answer.

01

Sourced results

Each memory refers back to where it came from.

02

Clear gaps

A response indicates when memory is empty or incomplete for the request.

03

Independent of your model

Recall works the same whichever model your product uses.

Where it fits / 03

For products whose users come back.

01

Assistants and copilots

Carry preferences, decisions, and open work across sessions.

02

Agent platforms

Give each run the user context it needs without replaying full histories.

03

Multi-model products

Switch or combine model providers without losing what each user has told you.

Access terms, supported regions, data handling, rate limits, and pricing will be published in formal documentation when the service is ready.

Early access

Building a product that needs memory?

Tell us what you are building. We will contact teams as access opens.

Contact us The memory model