July 26, 2026SkillsAgentsOpen Source

ECC: one config to run Claude Code, Cursor and Codex

Most people set up Claude Code once, drop in a CLAUDE.md, and call it done. ECC treats the harness like a product you actually engineer. It's a single system that ships 67 specialized subagents, 279 skills, a stack of hooks, coding rules for a dozen language ecosystems, and MCP wiring, and it runs the same across Claude Code, Cursor, Codex and OpenCode. One line to install: /plugin install ecc@ecc.

The bet is simple. The gap between a mediocre agent session and a great one usually isn't the model, it's the scaffolding around it, which skills it can reach, which subagent it delegates to, what fires when you save a file. ECC packages all of that so you're not rebuilding it from scratch in every repo and every tool.

It's trending hard on GitHub right now, and the reason isn't novelty. It's that the harness layer has quietly become where the real leverage lives. Once the model is good enough, the winner is whoever hands it the best tools, memory and instincts. ECC is a whole opinionated stack of exactly that, and it doesn't lock you to one vendor.

Worth a look if you're tired of copy-pasting your Claude setup between machines and tools: https://github.com/affaan-m/ECC
← Previous
Ops Log: July 25, 2026
Next β†’
Anthropic just deleted 80% of Claude Code's system prompt
← Back to all articles

Comments

Loading...
>_