mythos

The leaked Anthropic reasoning protocol. Running locally. Zero-drift coding with Strict Write Discipline and adaptive Claude Opus 5 thinking. Mythos

Claude Mythos Engine 2026: AI Sub-Agent Framework for Epic Interactive Storytelling

Claude 3.5 Sonnet Desktop Client for Windows Mac Linux 2026

This is an attempt to make original open mythos better with chat ui.

Outside-in replication of Anthropic's Mythos Preview / Project Glasswing — open-source agentic vulnerability-discovery scaffold on Claude Opus 4.7. Eight-phase sink-guided pipeline, ~$1/run, OSS self-scan and coordinated disclosure.

High-fidelity Claude Fable 5 (Mythos) environment emulation and automated multi-agent jailbreak (Pack Hunt) research laboratory.

Master prompts to make Claude (and other AI models) review code for security, performance, accessibility, usability and quality — language-agnostic, with add-ons for 18+ stacks. Bilingual EN/ES.

Mythos-class autonomous vulnerability discovery from any strong coding model (Opus 4.7, GPT-5.5, DeepSeek V4, Gemini). Long-running, email-driven, self-distills skills, mutates its own code.

One prompt. An LLM agent builds a complete procedural 3D landscape inside the Unity Editor. Terrain, splat maps, water, props, lighting, clouds, and a horizon ring — all from a single message.

🚨 The leaked source code of Anthropic's Claude Code. Explore 512k+ lines of TypeScript revealing multi-agent swarms, MCP integrations, Bun runtime, and a React/Ink terminal UI.

Visualization dashboard for parsing Mythos Router MEMORY.md files and SWD protocols on localhost

OpenMythos Claude Skill: a lightweight add-on that makes Claude act like a senior engineer who deeply understands Kye Gomez’s PyTorch RDT repo. It knows architecture, invariants, patterns, and debugging workflows—helping you reason, extend, and troubleshoot without modifying or redistributing the code.

A finish-first agent loop for Claude Opus 4.7 — plans, executes, verifies, and ships.

AI for Anti-Malware Research

Retired August 2026. Scaffold for Opus 4.x-class agentic discipline; Claude 5 models no longer need it. Code kept for older models.

An open-source framework for task-based agent swarms with dynamic parallelization, routing, and execution topology.

A free and strong UCI chess engine written from scratch in Rust — bitboards, alpha-beta search, NNUE neural-network evaluation, GPU training pipeline, and a built-in browser UI to play against it.

Mythos-class autonomous smart contract security reasoning engine. Bloodhound maps protocol state-spaces, identifies complex logic gaps, and chains anomalies into verified exploit hypotheses using a model-agnostic LLM loop. Built for deep EVM auditing and agentic security workflows.

A curated list of autonomous AI agent frameworks, tools, benchmarks, and papers

A deterministic, instrumented experimental framework for studying cycle-stealing schedulers, recursive control, and Mythos candidate exploration in MATLAB.