ARGI [/ˈɑːr.dʒiː/]: Agent Runtime and Graph Intelligence 🚀
A best-practices template for AI-assisted development at scale. It turns your repository into institutional infrastructure where AI agents can work sustainably — converging around documentation as a stable attractor across sessions, long cycles, and multi-role collaboration
CLEAR Context OS is the maintenance layer your second brain never had: ownership boundaries enforced mechanically, staleness surfaced on a schedule, drift propagated across linked docs, a dispatcher that learns from your clicks. One source of truth per topic. Searchable. Task-aware.
Cut LLM context and Claude Code sessions 40–80% without rewriting a byte. Offline, verbatim, with an audit receipt for every block dropped. CLI, MCP server and Claude Code/Open AI Plugin. MIT.
Persistent computational understanding for the AI tools you already use. A governed cognitive layer that transforms distributed observations into reusable understanding across models and interfaces.
Declarative framework for orchestrating multi-model Agentic pipelines with context engineering and quality gates.
Hyperdrive DCF (Design Cascading Framework) — a top-down methodology for rapid POC generation and idea validation, with every change tracked and synced back to the requirements spec with precision. Validated requirements then cascade through architecture and module specs to generate fully traced, test-gated production code. Powered by Claude Code.
Knowledge as code for teams where agents implement: decisions and rules in markdown, approved by humans, checked in CI, and found by agents in seconds.
Context Engineering library for LLM pipelines. Compression, token budgeting, context modes, and provider-agnostic execution.
Turn AI coding chats into a source-controlled workflow: AGENTS.md, rules, skills, verification gates, and adapters for Codex, Claude Code, and Cursor.
Open source, free, private, local, generative co-intelligence scaffold
Make every agent session start smarter than the last. A Claude Code skill for end-of-session knowledge maintenance.
External control plane for coding agents. Delegate repository exploration and evidence gathering while keeping reasoning and implementation with the primary agent.
CRUMB is the copy-paste AI handoff format. Switch AIs without losing the plot.
Day 6: Memory Architectures - Context Engineering Workshop
An Agent-safe context compressor for DeepSeek Harness. It semantically preserves critical tool outputs such as errors, file paths, JSON payloads and command results, instead of mechanically truncating the context.
Local, deterministic MCP server for writing agents. Indexes InkOS/Markdown/TXT/EPUB into a rebuildable SQLite/FTS5 index and answers with SHA-256-verifiable evidence excerpts under token budgets. No LLM inside, never writes back.
Plain-file task specs that coding agents can follow phase by phase: a stable plan, one brief per phase, and a verified handoff after each.
Interactive RAG Context Ingestion & Chunking Diagnostics Workbench evaluating semantic coherence via Gemini 3.8 Flash.
Explicit-context synthesis for Obsidian: select exactly what the model sees, preserve source snapshots, and synthesize with OpenAI.
To associate your repository with the context-engineering-framework topic, visit your repo's landing page and select "manage topics."