Skip to main content
Kodelyth ECC
Documentation

Every doc, every reference

Complete reference for Kodelyth ECC. Content synced live from the sifxprime/kodelyth-ecc GitHub repo.

Features
Codebase Graph — AST Code Intelligence Across 158 Languages
Kodelyth ECC wires DeusData codebase-memory-mcp for AST-parsed knowledge graph across 158 languages. Structural queries at 99% fewer tokens than file-by-file grep.
External MCP Servers — Register Stripe, GitHub, Postgres, Redis
Register external MCP servers with Kodelyth ECC. Auto-detect tools + prompts + resources from Stripe, GitHub, Postgres, Redis, Slack, and any MCP-compatible service.
Intent Routing v2 — Plain-Language to Specialist Agent in Claude Code
How Kodelyth ECC routes plain-language messages to 70 specialist AI agents automatically. 8 confidence dimensions, session-state awareness, terse-mode integration, evolve pipeline.
Interactive CLI — Arrow-Key Menu, Update Check, Background Daemon
Type kodelythecc alone in a terminal for an arrow-key menu with live update check, dashboard, IDE installer, and background daemon. Zero-dep raw-mode CLI in Kodelyth ECC.
Local Observability Dashboard — Kodelyth ECC
Localhost-only observability dashboard for Kodelyth ECC — Memory (BM25), RTK savings, Terse mode, Codebase graph, Arena, Evolve, Catalog, Sessions. Zero telemetry, zero external deps.
MCP Server — Universal Adapter for AI Agents (Kodelyth ECC)
ECC MCP server exposes 71 agents, 196 skills, 104 commands, 15 rules, BM25 memory to any MCP-compatible client — Claude Desktop, LangGraph, AutoGen, CrewAI, OpenAI Agents SDK.
RTK Integration — 60-90% Input Token Savings for Claude Code and Every AI IDE
How Kodelyth ECC wires RTK (Rust Token Killer) into Claude Code, Cursor, Windsurf, and every supported AI IDE for 60-90% shell-command input token savings — automatically.
Self-Evolving Memory — Compound Learning in Kodelyth ECC
Turn repeated memory hits and routing misses into PR-ready skill upgrades and rule additions. Never auto-applies — produces drafts you review.
Session Replay — Deterministic AI Session Bundles (Kodelyth ECC)
Export any Kodelyth ECC coordination session as a portable bundle. Deterministic replay with variations. Perfect for reproducing bugs and A/B testing agent workflows.
Super Agent — the Autonomous Orchestrator (Kodelyth ECC)
Drive one large objective to a verified result through ECC's Super Agent: a deterministic, resumable, budget-governed state machine that decomposes into a dependency DAG, dispatches independent steps in parallel on their own branches, conflict-checks and integrates, then verifies. Nothing faked.
Supply Chain Verification — SBOM, Manifest, SLSA L3 Provenance
CycloneDX 1.5 SBOM, sha256 content manifest, SLSA L3 provenance verification for Kodelyth ECC. Verify your installed toolkit hasn't been tampered with.
Swarm Orchestrator — Parallel AI Agents in Git Worktrees
Run N specialist AI agents simultaneously in isolated git worktrees + tmux sessions. Auto-picks agents from task signals. Maximum parallelism for Kodelyth ECC.
Terse Mode — 40-70% Output Token Savings for AI Coding Assistants
Compress AI reply length by 40-70% without losing information. 4-level dial, byte-preserves code and commands. Kodelyth ECC's answer to output token cost inspired by Caveman.
The Arena — GOD vs EVIL Adversarial Loop in Kodelyth ECC
Two agent crews fight over your code until the attacker gives up. GOD builds and hardens, EVIL attacks and proves, verified findings return to GOD. Scored, budget-capped, and it remembers what it learned.
pillar
Autonomous Agent Orchestration — The 2026 Reference
Complete reference for autonomous agent orchestration: the five hard problems (decomposition, dispatch, integration, verification, budget), the tokenless state machine pattern, dependency DAGs with cycle detection, git merge-tree conflict checking, resumable runs, and honest comparison to LangGraph, AutoGen, CrewAI, and OpenHands.
Codebase AST Knowledge Graphs — The 2026 Reference
Complete reference for codebase AST knowledge graphs: how tree-sitter builds them across 158 languages, why structural queries beat grep by 99% on token cost, the three queries that cover most practical use, Hybrid LSP semantic types, honest limits (staleness, type-resolution coverage), and when grep is still right.
How to Cut Claude Code Token Costs — Measured, Not Guessed
Where tokens actually go in an AI coding session, and the three levers that move the bill: filtering tool output before it reaches the model, compressing replies, and matching model tier to task. With a real 1,285-command ledger showing 64.1% input reduction.
Model Context Protocol (MCP) Server Guide — The 2026 Reference
A complete reference for Model Context Protocol (MCP) servers: what MCP is, how to build one, how to consume one, the six IDEs that auto-wire today, honest tradeoffs vs plugin ecosystems, security + transport patterns, and when MCP is the wrong answer.
The Claude Code Toolkit — Agents, Skills, Commands, Hooks, Rules, MCP (2026)
Complete reference for building a toolkit around Claude Code: the six primitives Anthropic exposes (agents, skills, slash commands, hooks, rules, MCP), the directory layout, how settings.json composes, intent routing, Memory Core integration, token-savings wiring, and when to extend each primitive.
comparison
Kodelyth ECC vs Aider — Augment the Pair-Programmer, Don't Replace It
Aider is a terminal-based AI pair-programmer with first-class git integration. Kodelyth ECC is a toolkit that augments pair-programmers with 71 specialist agents, MCP server, memory, and token savings. Install ECC into Aider rather than choose between them.
Kodelyth ECC vs Cline — Two Different Layers, Not Competitors
Cline is an AI agent harness that lives in VS Code. Kodelyth ECC is a toolkit that augments agent harnesses — including Cline — with 71 specialist agents, MCP server, memory, and token savings. Use both, not one or the other.
Kodelyth ECC vs Continue.dev — Config-Driven Assistant vs Catalog Toolkit
Continue.dev is a config-driven AI coding assistant for VS Code + JetBrains with first-class local model support. Kodelyth ECC is a catalog toolkit of 71 specialist agents + MCP + token savings. Honest comparison of the layers, with the integration path spelled out.
Kodelyth ECC vs Cursor Rules — When to Use Which (and Why Both)
Cursor Rules is a flat text file of always-on instructions. Kodelyth ECC is 71 specialist agents + MCP server + token savings + memory. They solve different problems at different layers — and ECC installs cleanly into Cursor alongside your existing rules.
Kodelyth ECC vs OpenHands — Autonomous Agent Harness vs Specialist Toolkit
OpenHands (formerly OpenDevin) is an open-source autonomous software engineer that owns its own agent loop. Kodelyth ECC is a toolkit that augments agent harnesses in your IDE. They solve different problems; this page is honest about whether they can be used together today.
Kodelyth ECC vs Plain Claude Code — Should You Add a Toolkit?
Plain Claude Code is a complete AI coding environment. Kodelyth ECC is a free toolkit that layers 71 specialist agents, a universal MCP server, local memory, and 55-65% token savings on top — without replacing what Claude Code already does well.