Context Kit — solve AI context amnesia with Personal Context Artifacts

[JDDavenport]

Hi — sharing a tool I built that addresses a gap across most AI tools: persistent personal context.

Context Kit — MIT, one-command install.

The problem

Every AI session starts at zero. The model doesn't know who you are, what you're building, or how you think. No matter how much context you've built up, the next conversation starts from scratch.

The solution: Personal Context Artifacts (PCAs)

4 Markdown files that capture you:

~/.claude/context/
  pca-wiki.md          # Who you are: roles, projects, goals, context
  pca-mental-models.md # How you decide: risk/time/money priors
  pca-voice.md         # How you write: 10 examples + anti-examples
  pca-protocols.md     # Your rules: non-negotiables, always/never

You fill them out once. They work with any AI tool, any model, any provider. You own them — they're just Markdown files.

5 Claude Code skills included

• /open-loops — surfaces open promises before they're dropped
• /watchers — sets up conditional monitors (file, email, URL)
• /crm-everything — logs every person mentioned
• /morning-briefing — context-aware daily briefing
• /session-digest — continuity notes between sessions

Install

curl -fsSL https://raw.githubusercontent.com/JDDavenport/context-kit/main/install.sh | bash

GitHub: https://github.com/JDDavenport/context-kit
Full essay (how I built a 28-agent AI system around this): https://docs.agenttree.army/articles/personal-ai-os/

The underlying pattern (static Markdown context files) works with any AI tool. Happy to discuss implementation or extensions.