SkillFoxx

daidocs MCP

daidocs keeps assistant memory as plain .dai files on your disk and connects it over MCP to Claude Code, Cursor and Codex.

  • 55 ★ on the repository
  • Last commit
  • Apache-2.0
MCP server

Extended access

Why this level

  • The installer edits several clients' configs and installs session hooks, and the memory saves the content of your conversations into files on disk.
All reasons and checks

kerneta/daidocsAuthor:Kerneta

Install

Manual install

npx daidocs setup

The first command of the Install section. For the `.cai` tier use `pip install kerneta-cai` separately.

Access and checks

How we review

Extended accessThe tool runs code or makes network calls and only reads external data.

  • The installer edits several clients' configs and installs session hooks, and the memory saves the content of your conversations into files on disk.

This is third-party code. Review the repository files before installing.

What it does

daidocs is a memory format for AI and an MCP server that reads and writes it. Conversation history and documents live in `.dai` files of three plain-text zones: a YAML header, a JSON block and text. Editors, `grep` and `git` open them, and Claude, GPT, Gemini and local models read them. A `.cai` code graph tier can sit beside it, answering who calls a function and what a rename touches. `npx daidocs setup` detects Claude Desktop, Claude Code, Cursor, Windsurf, Codex, Cline, Continue and Zed, configures them, installs session hooks and a reading protocol and backs up every file it touches. The setup can be undone with `node setup.js --restore`.

Who it is for. People who want assistant memory in their own files, not a vendor service, readable by any model and ordinary tools.

Good fit when

  • You want Claude Code to remember earlier sessions of a project even after a restart.
  • Conversation history must be kept in plain files and checked with `grep`.
  • One memory set must work in Cursor, Claude Code and Codex at once.
  • A call graph over code is needed to judge the impact of a rename.

Not a fit when

  • Cloud memory with a web UI and sync between machines is needed.
  • The installer must not edit client configs: it changes them though it backs them up.
  • A notes file that you update by hand is enough.

Example request

Recall from daidocs memory what we decided about the auth scheme last time and remember the new decision on tokens.

MCP

Transport
stdio
Authentication
not required

Limitations

Node 18 or newer is needed, and for the `.cai` tier Python 3.10 and `pip install kerneta-cai`. On a Claude subscription no key is needed because the assistant itself does the extraction; other models need a provider key. The installer edits found client configs and installs hooks, with backups kept beside them. The LongMemEval results in the README were measured by the authors and come with verification files. The last commit was in October 2026.

How to disable. Run `node setup.js --status` to see what is on and `node setup.js --restore` to return the machine to its earlier state. `.dai` files in projects stay as plain text.

FAQ

What is a .dai file?

A text file with three zones: a YAML header, a JSON block and text. You can read it in an editor or with `grep`, no special program needed.

Which clients get configured?

Claude Desktop, Claude Code, Cursor, Windsurf, Codex, Cline, Continue and Zed. For a single client use `node setup.js --client codex` and similar.

Is an API key needed?

Not on a Claude subscription: the assistant writes the extraction itself. Other models need a provider key.

How do I undo the install?

Run `node setup.js --restore`: each touched setting returns from its backup. `node setup.js --status` shows the state.

What is the .cai tier for?

It is a code graph answering who calls a function and what a rename touches. It installs separately with `pip install kerneta-cai`.

Discussion

No questions or reviews yet.

MCP servers

Local memory for Claude Code, Codex and Claude Desktop: one Markdown folder and one MCP server shared by all.

MCP serverExtended access103

hipocampus gives an agent file-based memory: a 3,000-token topic index loads each session and shows what came up before

PluginExtended access211

A guide with ready files: Claude Code keeps decisions between sessions through an Obsidian vault and /save and /resume commands

WorkflowBroad access1.1K

Shared long-term agent memory: notes in plain markdown files, with a MEMORY.md index loaded at the start of every session.

SkillExtended access204

Persistent memory for coding agents: insights are stored as markdown cards and recalled at the start of the next session.

MCP serverExtended access144

README in short

Explains the `.dai` format and the `.cai` tier, model independence, install paths, supported clients and benchmark results with verification files.

README badge

Report a problem