Add OpenClacky to Open-Source Projects — token-efficient LLM-powered coding agent

[streetlightstartupnotes]

What's being added

OpenClacky — an open-source LLM-powered autonomous coding agent.

GitHub: https://github.com/clacky-ai/open-clacky

Why it fits Autonomous Task Solver Projects

OpenClacky is a general-purpose AI agent that autonomously plans, codes, browses, runs commands, and manages files — in the same family as Auto-GPT, GPT Engineer, and MetaGPT, but specifically optimized for coding workflows and production deployment.

What makes it notable

Prompt Cache efficiency is OpenClacky's core technical contribution to the LLM agent space:

• 93.8% cache hit rate (7-day global average) — achieved by keeping the system prompt frozen and using dual cache markers
• ~0.8× the cost of Claude Code per session — via Insert-then-Compress context management
• This is a measurable, reproducible result, not a marketing claim — the cache_read_input_tokens are returned directly from Anthropic's API

Architecture: 16 stable tools, Skill-based workflow extensions, BYOK (Claude/GPT/DeepSeek/Kimi/Gemini/OpenRouter), IM integration, browser automation via Chrome DevTools, scheduled tasks via cron.

Website: https://openclacky.com
GitHub: https://github.com/clacky-ai/open-clacky
License: MIT