Memory

Overview

Memory is free-form text Kenny reads on every request to steer how it responds. Three layers stack together — your personal memory, your workspace’s memory, and each project’s memory — so you can pin preferences, conventions, and project context once and have Kenny apply them automatically across the chat drawer, research workspace, document assistant, and comment replies.

Use it for: setting tone and format preferences, capturing house style, recording project goals and conventions, pointing Kenny at documents it should always consult, and reducing repeated instructions in every prompt.

How it works

  1. Each memory layer is just a markdown note that you (or a workspace owner, or a project manager) write.
  2. Kenny reads all three layers together at the start of every request — personal first, then workspace, then project — so more specific notes can refine or override broader ones.
  3. Memory is private to the surface that owns it: your personal memory is only used when Kenny is responding to you; workspace memory applies to everyone in the workspace; project memory applies whenever Kenny is acting in that project’s context.
  4. Updates take effect immediately — the next prompt Kenny answers uses the latest memory.

Setup

Memory is optional. Kenny works without it. Add memory when you find yourself repeating the same instruction in multiple prompts, or when there are project conventions a new teammate would have to learn.

Usage

Edit your personal memory. Open Settings → Memory — or pick About you from your avatar menu to jump straight to your personal memory. Write free-form notes about how you like Kenny to talk to you — tone, formatting, defaults. Examples: “Default to bullet-point summaries with one-line headings.” “When I ask for code, prefer Python over JavaScript.”

Edit workspace memory. Open Workspace settings → Memory (owners can also use About workspace in the workspace switcher menu). Only workspace owners can edit it. Write things every member of the workspace should benefit from. Examples: “Our company tone is direct and friendly.” “When summarising a meeting, always end with an action items section.”

Edit project memory. Open Project settings → Memory. Project managers can edit it. Write project-specific context — what the project is for, who decides what, which knowledge documents Kenny should always consult, formatting conventions for tasks in this project. Examples: “New tasks go to the Backlog column at the bottom.” “Reference the Architecture document before answering technical questions.”

Combine layers. Personal preferences plus workspace conventions plus per-project specifics let you keep each layer short and focused.

Mention documents and folders in memory. Memory is just markdown. You can include links to knowledge documents and folders so Kenny knows where to look — the AI surfaces will resolve them when needed.

Edge cases & limits

  • Plain markdown. Memory is text. No formatting beyond standard markdown is required, and Kenny ignores anything it can’t make sense of.
  • No silent merging. The three layers are stacked, not merged — Kenny sees all three. Write more specific layers as refinements, not as replacements.
  • No PII protection. Anything you write in workspace or project memory is visible to everyone with access to that workspace or project. Don’t put secrets there.
  • Personal memory stays personal. Your personal memory is only used for requests you initiate. Other members never see it.
  • Memory shapes, doesn’t force. Memory is guidance, not a hard rule. Kenny will follow it most of the time but may deviate when a request explicitly contradicts it.
  • Project memory shows up over MCP. Project memory is returned to external AI clients that connect through the MCP Server, so they get the same context Kenny has.
This site uses Just the Docs, a documentation theme for Jekyll.