Your AI tools start cold. OrientKit injects your portable cognitive context into any conversation, so the model shows up already knowing you.
{ "identity": { "name": "Billy", "role": "builder" }, "cognitive_patterns": { "typo_density": "increases in deep flow", "thinking_style": "systems-first" }, "active_project": "orientkit" // your context, your file, your rules }
Cursor's rules file only works in Cursor. Switch tools, lose everything. Your preferences don't travel.
Claude Code's instruction file lives per-project. Different repo, different conversation, blank slate.
None of these files know how you think. They store project rules, not cognitive patterns. The model never learns you.
Define your ~/.dup.json once. Your identity, your cognitive patterns, your active projects. Follows the open DUP spec.
@dup in Copilot ChatThe extension reads your file and injects your full context block. The model arrives knowing your thinking style, your current project, your patterns.
Log session highlights directly to your DUP file. Your context grows with you. Inject into CLAUDE.md for Claude Code sessions.
Nothing is sent to a server. No telemetry. No account. The file lives on your disk, the extension reads it locally.
DUP is an open protocol. Any tool, any editor, any AI can implement it. No vendor lock-in, no walled gardens.
Other context files describe the project. DUP describes you. How you think, not just what you're building.
One file, every tool. Works with Copilot Chat today. Claude Code injection built in. The spec is designed for universal adoption.
OrientKit implements the Dynamic User Protocol. One portable file that makes every AI conversation feel like a continuation, not a cold start.