Maxxwell by Rindler
Writing

Rebuild a Monday-style board for agents

2026-11-18

You will have a Monday-style board view running directly in Maxxwell:


You will have a Monday-style board view running directly in Maxxwell:

You can retire most external boards and still keep live visibility into Claude, Codex, Cursor, or whatever agents you already run.

This is a practical Maxxwell board view tutorial. It assumes you already use Monday or something similar and want that same shape without another browser tab.

Prerequisites

Before step 1, you need:

If you want a deeper backdrop on how Maxxwell handles multi-agent fleets, see this related guide: how Maxxwell tracks every coding agent session in one pane.


1. Map your Monday board to Maxxwell concepts

Goal: Translate your existing board into sessions, status labels, and goal tags so you know what you’re recreating.

In Monday terms you probably have:

In Maxxwell:

Do this mapping explicitly once:

For example:

You’ll wire these into Maxxwell in the next steps.

Common failure: Trying to mirror every column from Monday. Start with status + goal tag. You can always lean on git branches and labels for the rest.


2. Start Maxxwell and create an orchestrator seat

Goal: Create the “board brain” - a single orchestrator session you talk to instead of twelve terminals.

Desktop

  1. Launch Maxxwell.
  2. Start a new session for the orchestrator seat.
  3. Paste a written brief, for example:
You are the orchestrator for the checkout-flow rewrite.

We track work with status labels:
- not started
- working
- waiting on you
- blocked
- done

We tag sessions with goals:
- checkout-flow
- auth-refactor
- infra-ci

Your job:
- Keep a running inventory of active sessions.
- For each session, maintain status and goal tag.
- Ask me only when a decision truly needs a human call.
- Report back periodically: what landed, what is still running, what is waiting on me.
  1. Start the orchestrator. It appears as its own session, just like a worker.

CLI

If you run the standalone CLI (names here are illustrative, adapt to actual binary names):

maxxwell orchestrator --name checkout-orch --brief ./briefs/checkout-orch.txt

Common failure: Treating the orchestrator as a dashboard you click rather than a conversation. It is a real agent seat; the more precise your brief, the better the board feels.


3. Spawn worker sessions for real tasks

Goal: Turn actual work into Maxxwell sessions so there’s something to put on the board.

Pick three to five tasks you’d normally create as Monday items.

Example tasks:

For each task:

  1. In Maxxwell, create a new worker session.
  2. Choose your tool, e.g. Claude Code, Codex, or cursor-agent.
  3. Give it a clear, one-paragraph prompt, for example:
Goal: Implement the order summary sidebar on the checkout page.

Constraints:
- Use existing tailwind utilities.
- Extend the current /checkout route.
- Add unit tests for the totals breakdown.
  1. Start the session. Maxxwell opens a real terminal session behind it.

Repeat for each task.

Now your “board” has rows. Each row is a live agent session you can attach to and take over at any time.

Common failure: Spawning one mega-session that does everything. Boards work when rows are small. If a task wouldn’t be a Monday item, don’t make it a session.


4. Apply status labels to each session

Goal: Recreate the Monday Status column by using Maxxwell’s status labels consistently.

Maxxwell already surfaces readable session states (e.g. working, idle, waiting on you, not heard from).

Agree on a minimal set, for example:

Then consistently use them:

  1. When you create a session, mark it not started until you actually hit run.
  2. When the agent is mid-task, keep it at working.
  3. When it asks you a question it cannot decide, mark waiting on you.
  4. When external blockers appear (missing secrets, failing services), use blocked.
  5. After you review and merge the change, mark done.

Mechanically, you can keep these labels in the orchestrator's brief and inventory, e.g.:

Set session "checkout-sidebar" status to working.

The orchestrator drafts the control rather than acting: it writes a sentence into the composer for you to send. You stay the one who presses enter.

Common failure: Overloading working as “has a long-running command” and “is stuck forever”. Use blocked or waiting on you so the board reflects reality.


5. Tag sessions with goals to emulate Monday groups

Goal: Use goal tags as your equivalent of Monday groups, epics, or sprints.

From step 1 you already picked a small set of goal names. Now actually use them — in each worker's prompt and in the orchestrator's inventory, so every session is attached to one goal.

Example goal tags:

For each worker session:

  1. Decide which goal it belongs to.
  2. Attach that goal tag.

Using the orchestrator, you can phrase it like:

Tag session "coupon-module" with goal checkout-flow.
Tag session "login-rate-limit" with goal auth-refactor.

Now, when you look at the Maxxwell fleet view, you can visually scan by goal, and ask the orchestrator to group its reports that way. That’s your Monday-style grouping.

Common failure: Letting goal tags explode into per-developer or per-minute labels. Keep them at the level you’d use for a Monday group - few, stable, meaningful.


6. Use the fleet view as your live board

Goal: Treat Maxxwell’s fleet surface as your Monday-style board: one pane that shows every session, its status, and its goal.

Once you have:

…open the fleet view. You now have something that behaves like a live board:

Because Maxxwell only shows what it can prove, you’ll see honest gaps like not heard from instead of fake working. Reducing one misaligned dashboard matters.

Common failure: Keeping Monday open “for safety” and drifting back into dual tracking. Pick one board-of-record per project. If Maxxwell is your live agent pane, let it own that.


7. Run your day off the orchestrator’s morning page

Goal: Replace your Monday “daily standup” scroll with a single orchestrator report.

Create a ritual: first thing in the morning, ask the orchestrator for a summary.

Example prompt:

Morning page request.

Summarize all sessions by goal and status.
For each goal:
- list sessions that are working
- list sessions that are waiting on me
- list sessions that are blocked
- list sessions that are done since yesterday

End with:
- three decisions I should make now
- one suggestion to prune or stop sessions

You now have:

This is the practical replacement for flipping through Monday boards.

Common failure: Letting the orchestrator become a passive log. Ask it specific questions and make it draft controls (e.g. “pause this session”, “start a new worker on X”) so your board stays writable.


8. Handle long-running sessions and context pressure

Goal: Keep your board readable as sessions accumulate and context bloats.

Maxxwell shows a live context-pressure readout per session with tiered warnings. Use it the way you’d use a WIP limit on a Kanban column.

Practical rules:

Use the one-click compact when needed:

Important: Maxxwell does not automatically compact or recycle context, restart work that stopped, or auto-correct drift. The conducting is real, the autopilot is not. Treat context warnings as prompts for you (or the orchestrator, via drafted commands) to act.

Common failure: Assuming sessions will auto-recover or self-prune. They won’t. That’s deliberate - you own the constraints.


9. Retire your external board safely

Goal: Move from Monday (or similar) to Maxxwell as your live board without losing track of work.

Do this in two passes.

Pass 1: Mirror

For a week:

Look for:

Fix the board that’s wrong. You’re testing whether Maxxwell can serve as the source of truth.

Pass 2: Cutover

Once the mirrored week feels sane:

At this point, the Maxxwell board view is your primary workflow surface for agents.

Common failure: Keeping “important” items as Monday-only reminders. If it matters, give it a session or at least a goal tag. A board with shadow tasks isn’t a board; it’s a lie.


FAQ / Troubleshooting

How do I recreate multiple Monday boards in Maxxwell?

Use different orchestrator briefs and goal tag sets per logical board.

Example:

You can run both orchestrators at once and switch between their views.

How do I track which agent session is stuck or stalled?

Maxxwell surfaces per-session states like working, idle, waiting on you, and not heard from.

When a session reads not heard from or has sat idle longer than you expect:

Can Maxxwell automatically re-aim a session that’s building the wrong thing?

No. Maxxwell does not auto-detect drift or re-aim sessions on its own.

What it does do:

You still decide when a session is off-goal and whether to kill, retarget, or re-prompt it.

How does Maxxwell compare to tools like Herd, Helmor, or ctx for this board use-case?

Maxxwell’s niche here is local, board-shaped visibility over the agents you already run, plus the orchestrator seat and the “drafts rather than acts” control model. It sits above Claude, Codex, Cursor, etc., instead of replacing them.

Does Maxxwell store my credentials or run anything in the cloud?

No. Maxxwell runs locally.

If you close your laptop, the agent sessions keep running the same way they would in your own shell.