Description
Cog is a memory system for AI agents, built on plain-text files. Instead of each coding tool keeping its own siloed memory, Cog gives you one structured, shared memory layer made of markdown files that any agent can read, search and maintain. Your preferences, project context and long-term goals live in one place and are shared across tools such as Claude Code, Codex, Cursor, Gemini CLI, Windsurf and Copilot. The site describes the idea as one memory rather than one per tool, so switching tools or starting a new project does not mean re-explaining yourself. Because the memory is plain text, it stays portable and easy to inspect and edit by hand. The project is described on its page with sections for architecture, getting started, the pipeline and the reasoning behind text as the format. Companion skills are published in a separate repository, cog-skills. The source is on GitHub under marciopuga/cog.…
Highlights
- One shared memory across AI coding tools
- Plain-text markdown files you can read and edit
- Works with Claude Code, Codex, Cursor and more
- Companion skills in a separate repository
Ratings & Reviews
No ratings yet
Be the first to rate Cog.
People also visit
Cog FAQ
What does Cog do?
It gives AI agents a shared, structured memory made of plain-text files, so context carries across tools.
Which tools does it work with?
The site lists Claude Code, Codex, Cursor, Gemini CLI, Windsurf, Copilot and more.
Why plain text?
The site explains that markdown files can be read, searched and maintained by any agent, and remain portable and easy to edit by hand.
Where is the code?
The main project is on GitHub under marciopuga/cog, and the companion skills are in marciopuga/cog-skills.
What are the best Cog alternatives?
Popular alternatives to Cog include Agent Memory, BRAIN.md and Website. See all Cog alternatives


