The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
Emulo
Mine your Claude Code and Codex logs into a local you.md agent profile.
Install
/plugin marketplace add ohad6k/emuloInstall command taken from the project's README or registry entry. Always review code before running it.
README excerpt
Read full README on GitHub →Emulo
Mine the messages you typed to your coding agents into a you.md profile, every rule with dated receipts, and install it where your agent can read it.
Your real coding-agent sessions already contain the rules you never wrote down: what “done” means, what you reject on sight, how you debug, how you design UI, and how you write when you are actually working.
Emulo mines selected evidence from those sessions (Claude Code, Codex, Copilot CLI, OpenCode, and Google Antigravity logs out of the box) into a private working profile, where every rule carries dated quotes from your own messages, and installs it where your agent can read it. The agent mining flow writes separate files for work, design, writing, and video, so a host can load only the one that fits the task.
The aim is an agent that works the way you do from the first message. That is an aim, not a result: what is tested so far is below, including the part that did not go our way.
Installing a profile and an agent actually reading it are two different things. In a host-by-host test on 2026-09-25, the profile reached the model through a project AGENTS.md in Codex (--target agents --repo ., verified end to end in one test) and was loaded into OpenCode's instructions. In Claude Code the you skill installs but is not reliably opened on its own; typing /you loads it. Cursor, Gemini, OpenClaw and Hermes Agent get the file where they look for it, and loading there is not verified. The support matrix has each one.
The Emulo Proof v1 methodology is an unexecuted methodology until a separately approved evidence release exists.
The video layer, and where it comes from
A spec commercial, made in Claude Design. No After Effects and no motion software: the whole thing is a composition rendered out to video.
It is spec work. It was not commissioned by Vercel and it was not made for them. A brand was picked to see how far the motion could go.
It is here because emulo:video is mined from sessions like the one that
produced it. The other layers work the same way: the profile is not a template,
it is what survived from real work.
Install
Inside Claude Code:
/plugin marketplace add ohad6k/emulo
/plugin install emulo@emulo
Inside Codex:
codex plugin marketplace add ohad6k/emulo --ref v0.6.7 --json
codex plugin add emulo@emulo --json
Then run emulo:mine and point it at your session history. Everything below explains what that produces and why. If you want the CLI instead of the plugin, see Quickstart.
Open source and privacy
Emulo is MIT licensed, free, and works without an account or a sign-in. The tool itself has nothing to buy. Separately, the site offers one optional paid service, a hand edit of a profile you already mined, and you never need it to use Emulo.
Session extraction, redaction, caches, the profile itself, and the agent adapters all stay on your machine. The one exception is mining: if you point it at a hosted model, the selected evidence goes to that provider. Point it at a local model and the whole run stays on your machine.
Not memory
Memory is what you explicitly told the model.
Similar plugs
Reverse engineer anything with agents, from app behavior down to native binaries.
Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype…
User-friendly AI Interface (Supports Ollama, OpenAI API, ...)
A cross-platform desktop All-in-One assistant for Claude Code, Codex, OpenCode, OpenClaw, Grok Build & Hermes Agent. Only official website: ccswitch.io