Every session starts at zero. Your Claude Code doesn't remember yesterday.
Open source Claude Code plugins that evolve with you.
Persistent memory, automatic context management, smart delegation. Install once - your system gets better every session.
the whole system, in 100 seconds
What changes in your terminal
One correction on Monday. A new session on Thursday that already knows it - and the five open source pieces that make that happen.
What the intro video shows
A correction typed on Monday - "You keep skipping tsconfig" - is repeated verbatim on Thursday because the session started at zero. Evolving Lite saves that correction the moment it is made. Kairn finds it again on Thursday from a differently worded question. Hooks block a tier 10 destructive command before it runs. PrimeLine Skills engage systematic debugging without being asked. The Planning Framework runs nine stages, from a DSV check through discovery, vehicle selection, the plan, interview, hardening, review, output and execution rules. Quantum Lens runs seven cognitive lenses - Void Reader, Paradox Hunter, Boundary Dissolver, Temporal Archaeologist, Scale Shifter, Failure Romantic and Interference Reader - built so that unanimous agreement counts as a failure state. The five pieces stack, each one optional. Install with a single git clone; no config and no API keys.
42
deep dives published
4,200+
Claude Code sessions
480
agent configurations
4 feedback loops
How my Claude Code plugins make your sessions smarter
Learn. Protect. Delegate. Heal. All automatic.
Learn Loop
Catch Claude repeating a mistake and the detector saves it as a reusable note. Recall can surface it later, so it lives in the system instead of your head.
Context Loop
Never lose work to context overflow. At 70% usage you get a warning, at 93% the system auto-saves your progress and creates a handoff.
Delegation Loop
Big searches and research tasks automatically route to specialized sub-agents. Your main context stays clean and focused.
Heal Loop
Every hook proves it works on every run. If a Claude Code update breaks something, you see the warning immediately - no silent failures.
open source ecosystem
Free Claude Code tools - every layer compounds
Start with the foundation. Add what you need, when you need it.
Each layer is optional. Each layer strengthens the one below.
$ one clone or one pip install, per tool
All free. All MIT licensed. No API keys. No build step.
the whole system, mapped
See how every layer connects
An interactive map of the open-source ecosystem: Evolving Lite, Kairn, UPF, Quantum Lens and PrimeLine Skills. Click any node for the deep dive.
prefer to watch?
The anatomy, in under 4 minutes
A guided tour of the system: the mind that decides, the reflexes that never ask, the memory that survives every session, and the self-healing layer - one moving part at a time.
Claude Code tutorials and deep dives
Patterns, plugins, and workflows for production Claude Code
Install in 30 seconds. No config. No API keys.
$ git clone https://github.com/primeline-ai/evolving-lite ~/.claude/skills/evolving-lite
# then: bash ~/.claude/skills/evolving-lite/setup.sh
Free. Open source. MIT licensed.
stay in the loop
Claude Code patterns, new tools, and build-in-public updates. No spam.
>_ Get the free Claude Code guide
>_ No spam. Unsubscribe anytime.

![Claude Code Too Verbose? Why Your Rules Stop Working [2026]](/_next/image?url=%2Fblog%2Fclaude-code-too-verbose-hero.webp&w=3840&q=75)
![Run Parallel Claude Code Sessions From One Terminal [2026]](/_next/image?url=%2Fblog%2Fparallel-claude-code-sessions-hero.webp&w=3840&q=75)
![Claude Code Skills vs Plugins vs MCP vs Subagents [2026]](/_next/image?url=%2Fblog%2Fskills-plugins-mcp-subagents-hero.webp&w=3840&q=75)