Skip to main content
A project is the top-level container for a piece of software you’re specifying. Each project has its own canvas, memory store, PRD history, and integrations.Projects belong to an organization and can be active or archived. Deleting a project is a soft delete — it can be restored.
Nodes are the items on your project canvas. Each node has a type (feature, data model, user flow, etc.), a title, and structured data — a flexible JSON object that holds the node’s details.Nodes also have position and size on the canvas, and can be collapsed to save space.
Edges are connections between nodes. They represent relationships — a feature depends on a data model, a user flow includes a feature, etc.Each edge has a source node, target node, and handle positions (top, bottom, left, right) that control where the connection attaches visually.
Memories are pieces of context stored per project — decisions you’ve made, things you’ve learned, progress updates, and preferences. Each memory has a category and a status.Memories are embedded as vectors using pgvector, enabling semantic search. When AI coding tools connect via MCP, they search these memories to understand your project context.Kommit detects conflicts when new memories contradict existing ones, and prompts you to resolve them.
Templates are reusable project configurations. When you create a new project, you can apply a template to pre-populate the canvas with a standard set of nodes and structure.Admins manage the template library for the organization.
Organizations are the multi-tenant boundary in Kommit. Each org has its own projects, members, billing, and integrations.Members can have one of three roles: owner, admin, or member. Row-level security ensures data isolation between organizations at the database level.