Engramory skill
Shared long-term agent memory: notes in plain markdown files, with a MEMORY.md index loaded at the start of every session.
- 204 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- The write hook and Python scripts change note files and client settings.
- Notes may hold private project agreements and sit on disk as plain text.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/engramoryDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The write hook and Python scripts change note files and client settings.
- Notes may hold private project agreements and sit on disk as plain text.
This is third-party code. Review the repository files before installing.
What it does
Engramory stores agent memory as files. Each fact sits in its own markdown file with a type (for example `feedback`), dates and a rationale, while a length-limited `MEMORY.md` index holds one line per note. The index loads at session start and details open on demand. No database, vectors or server are needed, and files can be read and edited by hand. The project supports several clients. Claude Code gets a rule for `CLAUDE.md`, an optional `SKILL.md` and a hook that rejects writes over the limit. Codex uses `tools/engramory_init.py`, and DeepSeek Harness has a `dsh-engramory` plugin. One agent writes while others can read the same folder.
Who it is for. Developers using Claude Code or Codex who want project agreements kept in readable files rather than in someone else's database.
Good fit when
- The agent repeats the same mistake and the correction must be stored long term.
- Database migration agreements have to survive a session change.
- One memory should be read by both Claude Code and Codex from one folder.
- Notes must be reviewed by eye and kept in git next to the project.
Not a fit when
- You need semantic search across a large archive: there is no vector index here.
- You need server-side memory sync between machines without git.
- The environment forbids running Python 3.9 or newer scripts.
Example request
Remember as feedback: always run tests before reporting done, and add the note to the index.Limitations
For Claude Code installation is manual: the rule is pasted into `CLAUDE.md` and the hook is registered in `settings.json` from a ready snippet. The index limit protects context, but in clients without a hook it is enforced only by rules and the `tools/engramory_check.py` check. Documentation and sample notes are Chinese by default, with English in `README.en.md`. Only one agent should write a folder, otherwise notes conflict. The last commit dates to November 2026.
How to disable. Remove the rule from `CLAUDE.md`, unregister the hook in `settings.json` and delete the `.engramory-memory` folder if the notes are no longer needed.
FAQ
How do I connect Engramory to Claude Code?
Paste `rules-snippet.md` into `~/.claude/CLAUDE.md` or the project `CLAUDE.md`, optionally place the repository as an `engramory/` skill and register the hook from `hooks/settings.snippet.json`.
How do I connect it to Codex?
Run `python tools/engramory_init.py codex --project-root /path/to/project --install-skill`. The `--install-hooks` flag adds lifecycle reminders.
Where are the notes?
In the `.engramory-memory/` folder: the `MEMORY.md` index, one file per note and an `archive/` subfolder for outdated ones.
What is read-only mode?
One agent writes to the folder and another reads it. The `claude-reader` and `codex-reader` modes of `engramory_init.py` set up reading.
Is there a DeepSeek Harness plugin?
Yes, it is called `dsh-engramory` and installs with `dsh plugin --profile add dsh-engramory`.
Discussion
No questions or reviews yet.
Related
SkillsPiia Engram: portable memory for AI agents
Piia Engram
Local memory about you and your projects: Claude Code, Codex and Cursor read the same preferences and lessons from your disk
Long-term agent memory in local Markdown files with a link graph, semantic search and MCP access
Engram by Gentleman Programming
Engram (Gentleman Programming)
Persistent memory for AI coding agents in a single Go binary: an MCP server on SQLite with full-text search that remembers your project across sessions
hipocampus gives an agent file-based memory: a 3,000-token topic index loads each session and shows what came up before
Memory for AI agents with three types: facts, events and procedures that learn from failures; cloud by key or a local folder
Persistent memory for coding agents: insights are stored as markdown cards and recalled at the start of the next session.
README in short
The README shows the memory folder layout and a sample feedback note, then walks through setup for dsh, Claude Code, Codex and other clients. A separate section covers the one-writer, many-readers scheme.