Running several projects

Running three or five things at once is where “I’ll remember” stops working. It’s also where tuckit asks a question it doesn’t ask a single-project user: is this all one board, or several?

There’s no setting for it. It follows from two facts, so those come first.

The two facts that decide it

Card numbers count per workspace. The 42 in ACME-42 counts inside one workspace and nowhere else. Two workspaces both have a 42, and they’re unrelated cards with different prefixes.

One agent connection reaches exactly one workspace. When your agent connects, the browser asks which organisation the connection is for, and what it gets back works only for that one. An agent connected to your blog’s board can’t see your SaaS board, and can’t accidentally file something into it.

Everything below follows from those two.

Shape A: one workspace, an area per project

Your acme workspace holds areas called blog, saas, client-work.

• One board, one Inbox. Everything from every project lands in the same place and you sort it in one pass.
• One connection. Set it up once and every folder uses it.
• One run of numbers. ACME-1 through ACME-400, across everything.
• Cards move between projects in a click, because moving between areas is free.

This is right when the projects share a what should I do next question, when you genuinely sit down and pick the most important thing across all of them. Side projects for one business usually work like this.

It goes wrong when they don’t. Five unrelated projects on one board means the board can’t answer “what’s next” until you filter it, and a board you filter before reading is data, not a surface.

Shape B: one workspace per project

acme-blog, acme-saas, bigco-contract, each its own workspace.

• Boards stay readable. Each answers “what’s next” on its own.
• The numbers read like the project. BLOG-7 and SAAS-7 are obviously different things, which matters when you paste one into a chat.
• Separate Inboxes, so nothing lands in the wrong project.
• One connection per project, which is one of the two costs. See below.
• The other cost is money. On the hosted service, billing is per workspace: a second workspace is another $99 a month. Areas inside one workspace are free, which is why Shape A is the cheaper default in both senses.
• No cross-project view. There’s no page showing all your boards at once.

This is right when the projects have different clients, different timelines, or would never compete for the same afternoon.

Choosing

Ask which sentence you say more often:

• “What should I work on today?” → Shape A. You want one ranked surface.
• “What’s the state of the blog?” → Shape B. You think per project, and mixing them is noise.

If you’re unsure, start with Shape A. It’s the cheaper mistake: splitting later means making a workspace and re-creating a handful of cards, whereas merging later means reconciling two runs of numbers that already appear in old chat logs. It is also cheaper on the invoice, since one workspace bills once no matter how many areas and people are in it.

Adding a second workspace

Settings → Account → Organisations → New organisation. Pick a name and a short lowercase one for the address, and tuckit derives the card-number prefix from it. That’s the BLOG in BLOG-7. Your account can hold as many as you like, and the sidebar switcher moves between them.

Cards don’t move across workspaces. If you split a board later, re-create the cards you still care about and leave the rest.

Pointing each repository at the right board

With Shape A there’s nothing to do. One connection, installed once, works everywhere.

With Shape B, each folder needs its own connection — and each board has its own address, which is what keeps them apart:

https://app.tuckit.dev/mcp/acme-blog
https://app.tuckit.dev/mcp/acme-saas

Put your workspace’s short name after /mcp/. In Claude Code, run this inside the folder:

claude mcp add --transport http tuckit-blog https://app.tuckit.dev/mcp/acme-blog --scope project

--scope project writes it into that folder’s .claude/settings.json, so the server exists only there. On first use the browser opens, already showing the workspace the address named, and you approve it. Repeat per folder.

The addresses matter more than the names. A credential issued for one workspace is refused at another workspace’s address — not filtered, refused — so an agent in your blog folder cannot reach the SaaS board even if something goes wrong at the client end. That is the server’s rule and it does not depend on how your agent stores things.

Two addresses, both real

https://app.tuckit.dev/mcp still works and is what the plugin installs. It is the right one when you have a single workspace, and every connection made before workspace addresses existed keeps working unchanged.

The difference is direction: a credential for one workspace will not open another, and a credential from the plain address stays on the plain address. Narrowing is a new authorisation, which is why adding a second board sends you through the browser again.

If you use the plugin rather than a hand-added server, it installs one named tuckit on the plain address. For a second board, add a per-folder server by hand as above.

Two habits that only matter at this scale

Write things down anywhere, sort them in one pass. With several boards you’ll want to file everything perfectly the moment you notice it. Don’t. That’s the tax that makes you stop noticing. Leave things in the Inbox with no area and clear them in one sitting. Writing down without sorting is a proper path precisely so noticing stays cheap.

Write back when you switch, not before you sleep. With one project you can reconstruct yesterday from memory. With five you can’t, and the moment you lose is the context switch, not the end of the day. Before you close a project, leave the note and record what you saw. With the plugin installed, “reconcile the board” does it for you. The daily loop is the full version.

Where to go

• The daily loop: the day-to-day.
• What’s on the screen: what a workspace, area and card are.
• Connect your agent: the other AIs, and connecting by hand.