context-kit

agent
Security Audit
Warn
Health Warn
  • No license — Repository has no license file
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 5 GitHub stars
Code Pass
  • Code scan — Scanned 4 files during light audit, no dangerous patterns found
Permissions Pass
  • Permissions — No dangerous permissions requested

No AI report is available for this listing yet.

SUMMARY

4 PCA templates + 5 Claude Code skills — give any AI agent deep personal context. One-command install.

README.md

Context Kit — Your Personal AI Operating System

GitHub stars
License: MIT
Claude Code

Stop re-explaining yourself to AI every single session. 4 Markdown files that give Claude permanent personal context — who you are, how you decide, how you write, and your non-negotiables.

Context Kit gives Claude Code a persistent mental model of you. 4 PCA templates + 5 skills. 100% free and open-source.

# Install in 60 seconds
bash <(curl -fsSL https://raw.githubusercontent.com/JDDavenport/context-kit/main/scripts/install.sh)

Read the essay: I ran my life on a fleet of AI agents for 6 months →

If this saves you time, a ⭐ star helps others find it.Star on GitHub


What's included

templates/
  pca-wiki.md           — who you are (roles, background, relationships, domains)
  pca-mental-models.md  — how you decide (money, time, risk, energy priors)
  pca-voice.md          — how you write (examples, anti-examples, channel rules)
  pca-protocols.md      — your hard rules (non-negotiables, time rules, comms rules)
  domain-state.md       — per-domain state file convention
  CLAUDE.md.template    — starter CLAUDE.md that loads your PCA files

skills/
  crm-everything.md     — never forget a person: create/update contact files
  open-loops.md         — capture & track promises before they become dropped balls
  watchers.md           — durable "notify me when X" conditions via Claude
  morning-briefing.md   — generate a grounded morning status from your state files
  session-digest.md     — end-of-session summary that feeds the next session

Install (2 minutes)

# Option A — one-liner (recommended)
bash <(curl -fsSL https://raw.githubusercontent.com/JDDavenport/context-kit/main/scripts/install.sh)

# Option B — clone and run
git clone https://github.com/JDDavenport/context-kit.git
cd context-kit
bash scripts/install.sh

The installer:

  1. Copies PCA templates to ~/.claude/context/ (or your chosen path)
  2. Installs skills to ~/.claude/skills/
  3. Prints next steps for filling in your templates

MCP Server (Claude Desktop + any MCP client)

Context Kit includes a zero-dependency MCP server that exposes your PCA files as resources to any MCP-compatible AI client.

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "context-kit": {
      "command": "python3",
      "args": ["/path/to/context-kit/mcp_server.py"]
    }
  }
}

Your PCA files then appear as resources in Claude Desktop (and any other MCP client) — accessible without Claude Code.


The 4 PCA Files

Personal Context Artifacts (PCAs) answer 4 questions every AI agent needs to do its job well:

File Question What goes in it
pca-wiki.md Who am I? Background, family, domains, goals, current life-state
pca-mental-models.md How do I decide? Money/time/risk priors, decision frameworks
pca-voice.md How do I write? Writing examples, anti-examples, channel-specific rules
pca-protocols.md What are my hard rules? Non-negotiables that agents must never violate

Once filled in, reference them in your CLAUDE.md so every Claude session loads them:

## Personal Context
Read these files at session start:
- `~/.claude/context/pca-wiki.md` — who I am
- `~/.claude/context/pca-mental-models.md` — how I decide
- `~/.claude/context/pca-voice.md` — how I write
- `~/.claude/context/pca-protocols.md` — my hard rules

You start context-full, not context-zero.


The 5 Skills

Skills are Claude Code slash commands. Install them, then invoke with /skill-name.

Skill Invoke What it does
CRM Everything /crm-everything Creates or updates a contact file for any person mentioned
Open Loops /open-loops Extracts promises/commitments from the current conversation
Watchers /watchers Sets up a "notify me when X" condition Claude checks on request
Morning Briefing /morning-briefing Generates a grounded morning status from your state files
Session Digest /session-digest End-of-session summary that primes the next session

Pricing

All templates and skills in this repo are free and MIT-licensed. The paid tier adds a curated setup guide, advanced examples, and priority support.

Tier What you get Price
Free All 4 PCA templates + 5 skills + CLAUDE.md scaffold — everything in this repo Free
Setup Pack Step-by-step setup guide + advanced PCA examples + 30-day email support $39 →
Course "Build Your Personal AI OS" — 6-module video course: PCA, agent fleet, cockpit, life OS $97 pre-order →
Sponsorware Early access to new skills + private build log $10/mo

Who this is for

You're already running Claude Code. You've gotten it to do impressive things. But every session you're re-explaining yourself — your goals, your constraints, your voice, your context.

The commercial alternatives (Lindy, Claude.ai, the OpenAI agents) are building the same thing but they keep the context locked in their walled garden. Switch providers — start over.

These are Markdown files. They're yours. They work with any model, any tool, forever.

Before Context Kit: Claude asks "what's the context here?" on every session.
After Context Kit: Claude opens already knowing your goals, constraints, voice, and rules.


Built from receipts

This system was extracted from 6 months of running a real agent fleet:

  • 133 projects managed by AI agents
  • 28 domain agents covering health, family, work, school, investing, consulting
  • 47+ cron jobs — autonomous, recurring work
  • 203 automation scripts

Not a thought experiment. A distillation of what actually worked.


Contributing

Open a Discussion to share your PCA setup, suggest new skills, or ask questions.

Found this useful? ⭐ Star this repo — it helps other Claude Code users find it.


Built by JD Davenport · Essay · AgentTree Hub

Reviews (0)

No results found