ReMe plugin
ReMe keeps agent memory in Markdown files with wikilinks and installs as a plugin for Claude Code, Codex and Qoder
- 3,796 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- Stores conversation content on disk, including images
- The plugin writes to memory automatically after each turn
Install
- Works in
Manual install
pip install "reme-ai[core]"The first install command from the README. Then the plugin for your host from `integrations`.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Stores conversation content on disk, including images
- The plugin writes to memory automatically after each turn
This is third-party code. Review the repository files before installing.
What it does
ReMe is a local-first, self-evolving knowledge base for AI agents. Conversations and resources become readable, editable, wikilinked Markdown files with metadata. Indexes are rebuilt while the files stay in your hands: you can open, edit, move and copy them with ordinary tools. Search combines BM25, optional embeddings and wikilink expansion, so the agent gets the relevant lines instead of the whole base. Claude Code, Codex and Qoder have plugins with automatic recall, batched recording of completed turns and MCP tools. Other CLI agents can use the `reme-memory` skill. It works through the CLI, HTTP, MCP or Python.
Who it is for. People who want several agents to share one on-disk memory that can also be read and edited by hand.
Good fit when
- Claude Code and Codex must remember the same project agreements.
- Knowledge from conversations accumulates into notes later opened as ordinary files.
- Agent context has grown and only the relevant lines are needed.
- Memory is kept in git or synced between machines as a Markdown folder.
- The link graph between notes is browsed in a browser through ReMe Studio.
Not a fit when
- One session is enough for the agent and long-term memory is not needed.
- Conversations must not be saved to disk even locally.
- You want hosted memory with a ready login and no service to run.
Example request
Find the database schema agreements in ReMe memory and append the decision we made today to the note.Limitations
The Claude Code, Codex and Qoder plugins work while the ReMe service is running: start it first, then install the plugin. Conversations are stored, including images with context, so make sure secrets do not land in memory. Building Studio statically needs Node.js 22.13 or newer. Optional decision models install through the separate `reme-ai[core,decision]` extra and require an external API key. Versions 0.2 and 0.3 live in separate branches.
How to disable. Uninstall the plugin from the agent and stop the ReMe service. The workspace folder with Markdown files remains and is deleted or moved by you.
FAQ
How do I install it?
Run `pip install "reme-ai[core]"`, start the ReMe service and install the plugin for your host. Instructions are in `integrations/claude_code`, `integrations/codex` and `integrations/qoder`.
What does memory as a file mean?
Durable memory is stored as plain Markdown with frontmatter and wikilinks. Files can be read and edited by hand and indexes can be rebuilt.
Which hosts are supported?
Plugins for Claude Code, Codex and Qoder, a provider for Hermes Agent and a skill for other CLIs. QwenPaw and DeepSeek Harness integrations also exist.
What is ReMe Studio?
A web interface for viewing and editing Markdown memory and the link graph. A demo opens in the browser with no installation.
How does search work?
Search joins BM25, optional embeddings and wikilink expansion and returns lines with context rather than whole files.
Discussion
No questions or reviews yet.
Related
PluginsShared memory for Claude Code, Codex, Cursor and other agents in Markdown files on disk, with retrieval that explains its results
Shared semantic memory for agents: conversations are saved as markdown and recalled by search across Claude Code, Codex and other agents
Local memory for AI agents in plain Markdown files: it keeps only what you approve and serves it through MCP
Long-term memory layer for AI agents: decisions and facts are stored as JSON next to the code and found again in a later session.
Persistent memory for coding agents: insights are stored as markdown cards and recalled at the start of the next session.
Dual-path memory for agents: a knowledge graph on your machine, search on demand and hints before you ask, across sessions.
README in short
The README calls ReMe a local-first self-evolving knowledge base. The integrations table lists how to connect memory for each host and what it gives.