npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

holistic

v0.3.0

Published

Repo memory, MCP handoffs, and session continuity for AI coding agents

Readme

Holistic

██╗  ██╗ ██████╗ ██╗     ██╗███████╗████████╗██╗ ██████╗
██║  ██║██╔═══██╗██║     ██║██╔════╝╚══██╔══╝██║██╔════╝
███████║██║   ██║██║     ██║███████╗   ██║   ██║██║     
██╔══██║██║   ██║██║     ██║╚════██║   ██║   ██║██║     
██║  ██║╚██████╔╝███████╗██║███████║   ██║   ██║╚██████╗
╚═╝  ╚═╝ ╚═════╝ ╚══════╝╚═╝╚══════╝   ╚═╝   ╚═╝ ╚═════╝

Your repo remembers, so your next agent doesn't have to guess.
Shared memory for AI agents, built into your repo.

One command. Every agent. Zero re-explaining. ✨

Holistic gives your AI agents shared memory inside the repo itself. When you switch from Claude to Codex to Gemini, the next agent can see what happened last time, what not to break, and what should happen next.


Get started in 30 seconds ⚡

Open your project repo in PowerShell, Terminal, Command Prompt, or whatever shell you normally use.

Requires Node.js 24+.

Run these two commands:

npm install -g holistic
holistic bootstrap

After that, open the repo in your agent app and use this startup prompt:

Before doing any other work, read AGENTS.md and HOLISTIC.md, recap the current state briefly, and ask me exactly one question: continue as planned, tweak the plan, or start something new.

That is enough to get the basic Holistic workflow working.

If you want the fuller install and setup details, jump to Quick start.


The problem 😵

If you use more than one AI coding assistant, the workflow usually falls apart:

  • 🔁 You re-explain the project every session.
  • 🐞 Bugs come back because the next agent does not know what was already fixed.
  • 🧠 Progress gets lost when context windows end.
  • 💥 Agents undo each other because there is no durable handoff.
  • 🌫️ It is hard to tell what is actually done.

Holistic fixes that by making the repo the source of truth.


What it feels like now 🌿

Run one setup command on a machine:

holistic bootstrap

Then daily use is mostly:

  1. Open the repo in Codex, Claude, or another supported app.
  2. Start a fresh session.
  3. Ask the agent to read AGENTS.md and HOLISTIC.md.
  4. Let Holistic carry continuity through checkpoints, handoffs, and repo memory.

Most days, you do not need to run npm start, keep a terminal process open, or manually re-brief the agent.

holistic bootstrap is a machine setup command, not just a repo setup command. By default it can install local startup helpers and configure Claude Desktop MCP on that machine.


How it works 🧭

holistic bootstrap
      ->
You open a repo in your agent app
      ->
The agent reads HOLISTIC.md and AGENTS.md
      ->
"Here's where we left off. Here's what's next. Continue as planned, tweak the plan, or start something new?"
      ->
Work happens
      ->
Holistic checkpoints and handoffs keep repo memory current
      ->
The next agent picks up without a long re-explanation

Quick start 🚀

Install 📦

Requires Node.js 24+.

npm install -g holistic

Then verify the CLI is available:

holistic --help

For contributors or local source installs:

git clone https://github.com/lweiss01/holistic.git
cd holistic
npm install
npm run build
npm pack
npm install -g ./holistic-*.tgz

For local development without a packaged tarball:

npm install
npm link

Set up a repo 🛠️

cd my-project
holistic bootstrap --remote origin
git add .gitattributes HOLISTIC.md AGENTS.md CLAUDE.md GEMINI.md HISTORY.md
git add .holistic/config.json .holistic/state.json
git add .holistic/context/
git commit -m "feat: add holistic"

By default, Holistic now syncs portable state through a hidden git ref (refs/holistic/state) to avoid GitHub branch noise.

Advanced overrides:

holistic bootstrap --state-ref refs/holistic/state
holistic bootstrap --state-branch holistic/state

If you want repo scaffolding without changing local desktop integrations or daemon startup on the current machine, use:

holistic bootstrap --install-daemon false --configure-mcp false

What to commit:

  • .gitattributes - Holistic-managed line-ending rules for portable files
  • .holistic/config.json - repo configuration
  • .holistic/state.json - current session state
  • .holistic/context/ - generated docs (history, regression watch, adapters)
  • .holistic/sessions/ - session history files

What NOT to commit:

  • .holistic/system/ - machine-local helper scripts and wrappers with absolute paths (already in .gitignore)

The portable repo memory is meant to be committed and synced. Machine-local helper scripts are generated for each machine and stay local.


Daily workflow 🔄

One-time machine setup:

  • Run holistic bootstrap.
  • By default it scaffolds repo files, installs hooks, sets up daemon startup, and configures supported integrations such as Claude Desktop MCP on the current machine.
  • If you only want repo files and hooks, use holistic bootstrap --install-daemon false --configure-mcp false.

Normal use:

  • Start a session in Codex, Claude, or another supported app.
  • Let the agent read the repo instructions and current handoff state.
  • Work normally.
  • Use explicit CLI commands only when you want to inspect state manually or force a checkpoint or handoff yourself.

Useful manual commands:

holistic status
holistic checkpoint --reason "..."
holistic handoff

If holistic is not on PATH in a given shell, every bootstrapped repo also has a repo-local fallback:

  • Windows: .\.holistic\system\holistic.cmd <command>
  • macOS/Linux: ./.holistic/system/holistic <command>

Regression protection 🛡️

When an agent fixes something delicate, lock it in:

holistic checkpoint \
  --fixed "login redirect loop" \
  --fix-files "src/auth.ts" \
  --fix-risk "changing redirect logic will re-introduce this"

Future agents will see that warning in the repo docs before they touch the risky area again.


Works with multiple agent apps 🤝

Holistic is model-agnostic. It works through repo files first, and can also expose a thin MCP server where supported.

| App | Reads | Startup experience | |---|---|---| | Claude Desktop | CLAUDE.md and repo docs | automatic plus MCP support | | Codex | AGENTS.md and repo docs | automatic | | Gemini / Antigravity | GEMINI.md and repo docs | automatic | | Other VS Code forks | AGENTS.md and repo docs | usually automatic | | Web tools | repo docs pasted manually | manual |


What lives in your repo 🗂️

my-project/
|- HOLISTIC.md
|- AGENTS.md
|- CLAUDE.md
|- GEMINI.md
|- HISTORY.md
`- .holistic/
   |- config.json
   |- state.json
   |- sessions/
   `- context/
      |- project-history.md
      |- regression-watch.md
      `- adapters/

The portable repo memory (config, state, context, sessions) is meant to be committed and synced. Machine-local helper scripts and repo-local CLI fallbacks under .holistic/system/ are generated for each machine and stay local (already in .gitignore).


Commands

| Command | What it does | |---|---| | holistic init | Base repo setup and scaffolding | | holistic bootstrap | One-step machine setup for repo files, hooks, and by-default local daemon/MCP integration setup | | holistic start --agent <name> | Opens a session and prints the ASCII banner plus recap | | holistic checkpoint --reason "..." | Saves progress and context | | holistic handoff | Ends a session with a handoff | | holistic status | Shows the current state | | holistic diff --from <id> --to <id> | Compares two sessions | | holistic serve | Runs the thin MCP server and prints a startup banner to stderr | | holistic watch | Foreground daemon mode for automatic checkpoints |

Non-interactive handoff

holistic handoff \
  --summary "Implemented OAuth flow and token storage" \
  --next "Wire up the refresh token endpoint" \
  --blocker "Need refresh token endpoint from backend team"

Architecture

Holistic is intentionally repo-first, not machine-first.

| Layer | Purpose | Portable? | |---|---|---| | Repo memory | Shared handoff, history, regression, and session state | Yes | | State ref | Cross-device distribution of Holistic state via git | Yes | | Local daemon | Passive capture on one machine | No |

That split is what makes Holistic work across tools and devices instead of only on one laptop.

MCP server mode

Holistic can run as a thin MCP server for agent-native workflows:

holistic serve

In normal use, Claude Desktop can launch this automatically after holistic bootstrap configures the MCP entry. You usually only run it manually for debugging.

When you do run holistic serve manually in a terminal, Holistic prints its ASCII startup banner to stderr so you get visible confirmation without corrupting the MCP stdout transport.

{
  "mcpServers": {
    "holistic": {
      "command": "holistic",
      "args": ["serve"],
      "env": {
        "HOLISTIC_REPO": "/path/to/your/project"
      }
    }
  }
}

Why this matters

If you are already using more than one AI coding assistant, you already have the continuity problem.

Holistic gives you:

  • Less repeated explanation
  • Fewer accidental regressions
  • Clearer handoffs across apps and devices
  • A durable record of what changed and why
  • Agents that can get to work quickly

Quick links