How does a Redeep project board work?

A Redeep board turns a familiar Kanban card into an active piece of work. Instead of merely reminding a person what to do, a card holds the full brief, conversation, working files, and progress of the AI worker assigned to it — a persistent work session that can read your files, analyze information, make changes, and browse the web, carrying one task to a clear finish.

Create the stages that fit your project — say Ideas, Ready, Working, Review, and Done. Write the desired outcome on a card. Move it into the working column and the task can begin. The card shows whether work is running, finished, or waiting on a decision from you.

I like that this makes the state of a project visible at a glance. You don't have to remember which chat has the pricing analysis, whether the presentation draft is finished, or where the website edits landed. The board is both the plan and the live workbench.

One card, one outcome. “Turn the interview notes into a two-page research summary” gives the worker a destination. The board gives that work a place to live.

Can several cards run at the same time?

Yes. Redeep can run multiple cards in parallel while keeping their work isolated. Each active card gets its own working copy of the project, so two workers can make progress without saving over each other.

Imagine a product launch. One card updates the website. A second reviews support notes and drafts an FAQ. A third analyzes the last quarter of customer feedback. All three can move at once, even when they share the same underlying project folder.

For software projects, this isolated copy is called a git worktree. You don't need to know git to benefit from it — the everyday analogy is simple: each worker gets a clean duplicate of the project and returns a marked-up set of changes. You decide what joins the main version.

That separation removes one of the biggest costs of delegation: coordinating who is allowed to touch what, and when. Parallel work becomes useful rather than fragile.

How do column automations save time?

A column automation is an instruction that runs when a card reaches a stage. It turns a repeated handoff — “now do the review” — into a dependable part of the workflow.

I could ask every card entering Review to check claims against sources, summarize the changes, and list anything that needs a human decision. A publishing board could optimize the title and description when a draft reaches Ready to Publish. An engineering board could run the relevant tests before work is presented for review.

Some stages shouldn't fire automatically. Redeep also lets you attach an action that waits for you to press play — useful when cost, timing, or judgment matters. You can even run one instruction across every card in a column, like updating a group of launch tasks after a date changes.

The result is not automation for its own sake. It is fewer dropped steps, more consistent review, and less time repeating the same direction.

Can you hand a difficult card to another AI?

Yes. A card can be handed to Codex or Claude Code when you want a specialist or a second opinion. Comment with the appropriate handoff, and Redeep sends the task to the command-line tool already connected to your own account.

The external worker receives an isolated workspace, completes its assignment there, and returns a report plus changes you can inspect. Your provider credentials aren't passed along by Redeep. The work still belongs to the same card, so its history doesn't disappear into a separate window.

I reach for this when a task is unusually technical, when I want an independent review, or when a team has standardized on a particular tool for part of its process. You choose the right worker without giving up the board as the source of truth.

Do you have to design every workflow yourself?

No. Redeep includes templates for common kinds of work, and every board remains editable. Start with a useful structure, then rename columns, add instructions, or remove anything that doesn't fit.

  • Engineering: move from backlog through implementation, review, and completion with clean technical handoffs.
  • Data analysis: frame the question, gather and validate evidence, build the analysis, and review the conclusion before sharing it.
  • Marketing: carry ideas through research, drafting, editing, approval, and publishing.
  • Personal assistance: organize requests, follow-ups, scheduled work, and items that need your decision.

Templates are starting points, not rigid systems. A board should make the project easier to understand — if a stage doesn't help you decide or act, delete it.

For the full operating details, see the project boards documentation. Or explore the broader idea of building a crew around recurring roles.