Home

MEM (Markdown Embedded Memory) gives AI coding agents a project memory that outlives the chat. It is a knowledge base of plain Markdown files that lives in your repository. People review it in pull requests like any other file, and every agent session reads it before working and updates it afterwards.

MEM is a specification, not software. The core needs no runtime, no SDK and no dependencies: any agent that can read Markdown and write files in the repository can use it. That includes Claude Code, OpenAI Codex, GitHub Copilot, Cursor, Windsurf, and chat extensions in Visual Studio or VS Code. Optional extensions add commands and, where you allow it, small programs.

Why it exists

An agent session produces knowledge as it works: how a module really behaves, why a bug happened, which library you settled on, what is still pending. Without somewhere durable to put it, that knowledge stays in a conversation nobody will reopen. The next session, or a colleague's agent, has to work it out again, sometimes wrongly.

MEM gives that knowledge a predictable place: architecture notes, decisions (ADRs), conventions, tasks, troubleshooting runbooks and daily logs. It also gives the agent rules for when to read each part and when to write. The source code stays the authority on what the system does. MEM is the maintained explanation of how it works, why, and what changed.

Start in one prompt

Paste this into your coding agent at the root of your repository:

Download the latest MEM.md file from https://raw.githubusercontent.com/Ryadel/MEM/main/src/MEM.md
and place it under /MEM/MEM.md in this repository.

Then read and apply /MEM/MEM.md as the persistent operating memory for this project.
Initialize the MEM knowledge base if required,
then inspect the project before making implementation claims.

Getting Started covers what happens next, the manual install, and how updates work.

Pages

Page	What you will find
Getting Started	Install, first initialization, choosing extensions, updating
Knowledge Base	The folder structure, what goes where, what never goes in, ADRs, logs, tasks
How the Agent Works	Reading economically, delegating, temporary files, commits and credit, external actions
Configuration	Every MEM.config.md option with its default, precedence, remote MEM source
Commands	MEM <COMMAND>: grammar, core commands, extension commands, help requests
Extensions	How extensions work, are installed, updated and trusted; code-shipping extensions

Extensions

Extension	Version	In one line
mem-commands	1.0.0	MEM REVIEW, MEM CHECK and MEM DEFINE: review plans, verify "done" claims, write your own commands
mem-toolbox	1.0.3	Which tool or library to use for a task, and whether it is really installed on this machine
mem-workspace	1.0.0	The knowledge base in a private repository beside a public one, with an optional wiki
mem-stats	0.1.1	Measures what agent sessions consume (tokens, cache, context composition) into a local ledger
mem-optimize	0.1.2	Keeps long sessions from wasting tokens: guided restarts, long-pause guard, re-orientation after compaction
mem-cleaner	0.1.0	In development, not yet in the catalogue: cleaning pipelines over files agents write

Links

• Repository: https://github.com/Ryadel/MEM
• Specification: src/MEM.md
• Upgrade steps for agents: MEM.upgrade.md
• Extension catalogue: src/extensions/manifest.md

MEM is released under the MIT License.