Skip to content

how-to-use-continuum.md

Latest commit

 

History

History
177 lines (116 loc) · 7.87 KB

how-to-use-continuum.md

File metadata and controls

177 lines (116 loc) · 7.87 KB
Raw

How to Use Continuum (Plain English Guide)

This guide explains Continuum without the jargon. If you've never touched a command line before, or you just want a plain refresher, start here.

1. What is this thing, actually?

Think of Continuum as a diary for your AI conversations.

Every time you work with an AI — ChatGPT, Claude, Codex, whatever — you make decisions, hit dead ends, and figure things out. Normally, all of that disappears the moment you close the chat or open a new one. You end up re-explaining the same things over and over.

Continuum keeps a permanent, tamper-proof record of that work, organized by project, so that:

  • You (or a new AI session) can pick up exactly where you left off.
  • Nothing important — a decision, a constraint, a failed approach — gets forgotten.
  • Different projects stay completely separate, like separate notebooks.

One-sentence version: Continuum is a memory system that sits between your AI conversations so switching chats, tools, or days never means starting over.

2. The one idea to remember: Projects

Everything in Continuum lives inside a project. A project is just a named container — like a folder — for one piece of work.

  • Working on your website? That's one project.
  • Working on a mobile app? That's a different project.

Nothing crosses over between projects automatically. Project A's decisions never quietly show up in Project B's notes. If you're working on two different things, you should have two different projects.

There's always exactly one active project at a time — think of it as "which notebook is currently open on your desk." Every command you run acts on whichever project is currently active, until you switch it.

3. Setting up (one-time)

Before anything else, tell Continuum to set itself up on your machine:

continuum init

This creates a private folder on your computer (~/.continuum) where everything will be stored — nothing leaves your machine unless you explicitly export or share it.

Then double-check everything's healthy:

continuum health

That's it. You only do this once.

If continuum doesn't run as a bare word for you yet, use the full form instead: pnpm --silent dev:cli in place of continuum everywhere in this guide (run it from inside the project folder).

4. Your daily workflow

Here's what a normal day looks like, step by step, in plain words.

Step 1 — Make sure you're in the right project

Starting work on something? Either create a new project or switch to an existing one.

Starting something brand new:

continuum project new "Website Redesign" "Marketing site rebuild"

This creates the project and automatically makes it the active one — you don't need a separate step.

Going back to something you already started:

continuum project ls

This lists everything you've got, showing titles (not just cryptic IDs), and marks which one is currently active.

continuum project use "Website Redesign"

This switches the active project — you can type the project's name or its ID, whichever you remember.

Step 2 — Get your conversation into it

If you had a chat with an AI somewhere (ChatGPT, Claude, etc.) and want to save it:

continuum add ./my-chat.json

This reads the file, figures out on its own which AI it came from, and drops it into whichever project is currently active as a brand-new session.

Special case — Claude Code: if the conversation happened right here, in a Claude Code session, you don't need to export anything at all. Claude Code already saves its own transcript to your disk automatically. You can point Continuum straight at that file instead of manually exporting anything.

Step 3 — Jot things down as you go (optional)

You don't have to import a whole file every time. You can log things one at a time, like a running notebook:

continuum log note "Decided to use PostgreSQL instead of MongoDB"

Step 4 — Check what Continuum understood

Want to see what Continuum picked up from everything so far — the objectives, decisions, what's left to do?

continuum state status

This shows you a clean summary: goals, constraints, decisions made, what's done, what failed, and what's still open. Every single line traces back to the exact message it came from — nothing is made up or guessed without a source.

Step 5 — Hand it to a fresh AI session

This is the whole point of the tool. When you open a brand-new chat — different day, different tool, different model — instead of re-explaining everything, run:

continuum state resume

This prints a ready-made briefing document. Copy it, paste it into the new AI chat, and that AI now knows your goals, your constraints, your past decisions, and what's left to do — without you typing a single word of context yourself.

5. Everyday commands, explained like a human would say them

What you want to do What you type
"Set this whole thing up" continuum init
"Is everything working okay?" continuum health
"Start a new project" continuum project new "Name" "Description"
"Show me all my projects" continuum project ls
"Switch to a different project" continuum project use "Name or ID"
"Save this chat file" continuum add ./file.json
"Jot down a quick note" continuum log note "..."
"What has Continuum figured out so far?" continuum state status
"Give me a briefing for a new AI chat" continuum state resume
"Search everything for a keyword" continuum find "budget"
"Show me a timeline of everything that happened" continuum history
"Make sure nothing got corrupted" continuum check
"Show me a visual overview" continuum dash
"Save a portable backup of a project" continuum capsule backup
"Restore a project from a backup" continuum capsule restore ./file.ctx
"Record a decision I just made" continuum track decide add -c "what we chose" -r "why"
"Add something to my to-do list" continuum track task add -d "description"
"Mark a task as done" continuum track task update <id> -s completed
"Close out this work session" continuum session close

You don't need to memorize this table — you can also just describe what you want in plain English to an AI assistant sitting in your terminal (like this one), and it will run the right command for you.

6. Common mix-ups (and how to avoid them)

"I imported the wrong project's chat." This almost always means the wrong project was active at the time. Always run continuum project ls before importing anything, to double-check which project is currently active — it's marked with "← active" in the list.

"I switched projects but nothing changed." Make sure you actually ran continuum project use ... and not just continuum project ls (which only shows projects, it doesn't switch to one).

"The AI in my new chat doesn't seem to know anything." Make sure you actually pasted the output of continuum state resume into the new chat. Continuum only prepares the briefing — it doesn't automatically send it anywhere (yet).

"Where is all this data actually stored?" Everything lives in a hidden folder in your home directory: ~/.continuum. It's hidden by default — in Finder, press Cmd+Shift+. to reveal it. Each project gets its own subfolder, so nothing overlaps.

7. The short version

  1. continuum init — once, ever.
  2. continuum project new "..." — once per new piece of work.
  3. continuum add file.json or continuum log note "..." — as you go.
  4. continuum state status — whenever you want to see where things stand.
  5. continuum state resume — whenever you're starting a fresh AI chat and don't want to explain everything again.

That's the entire tool. Everything else is a convenience on top of these five steps.