Plan · Agents

Supervisor‑Worker Framework Visualizer

One supervisor splits a job across workers you choose, then reviews a share of what comes back. The page draws that path and counts the handoffs. It does not start agents, and it does not call a model.

Set workers, slice size, and review percent. An in-browser AI agent draws the path and counts handoffs and planning minutes. No agents start. No API key.

Three knobs, said in plain words

Workers. How many people, or agent roles, take a slice. The supervisor is extra. The supervisor is not one of these workers.
Slice size. Small is 10 minutes of work per worker. Medium is 25. Large is 45. Planning minutes, not a benchmark.
Review. The percent of worker notes the supervisor reads. Each reviewed note adds 5 minutes.

A worked example, before you move a slider

3 workers. Medium slices (25 minutes each). Supervisor reviews 40%.

Reviewed notes = ceiling of 3 × 40% = 2.
Handoffs = 1 brief + 3 returns + 2 reviews = 6.
Minutes = (3 × 25) + (2 × 5) = 85.
The path is: you, then the supervisor, then each worker, then the reviews, then one answer. Move a slider and the AI agent redraws it.

The squad you want to picture

Counts you can explain in a design review. Nothing here is running.

Workers 3 is the example.

Slice size Minutes of work you assign to each worker. Supervisor review (%) Share of worker notes the supervisor reads. 40% is the example.

40%

Handoffs, minutes, and the path

AI agent is ready. Run it, or move a slider.

Handoffs — Brief + returns + reviews

Reviewed notes — Workers times review percent, rounded up

Planning minutes — Worker slices plus review minutes

In plain words —

Path

The AI agent writes this after it runs.

A picture of a supervisor-worker split. Not a running cluster, and not a measure of a real team.

How the picture is counted

Reviewed notes = ceiling of workers × review percent.
Handoffs = 1 (the brief) + workers (each return) + reviewed notes.
Minutes = workers × slice minutes + reviewed notes × 5.
The path lists you, the supervisor, each worker, each review, then one combined answer.

The AI agent fills this line after it runs.

Unlock a printable summary

The path stays visible with no email. A work email unlocks Print / Save as PDF. Same subscribe flow as the newsletter.

FAQ

Does this start real agents?

No. It draws a path: you, one supervisor, the workers you set, the notes the supervisor reviews, then one answer. Nothing is executed.

How are handoffs counted?

One handoff for the brief, one for each worker return, and one for each note the supervisor reviews. Reviewed notes are the worker count times the review percent, rounded up.

What are the minutes?

Small slices are 10 minutes, medium 25, large 45. Each reviewed note adds 5 minutes. These are planning assumptions so a design review can see the shape of the work. They are not a benchmark.

Do I need an API key?

No API key is required. No OpenAI, Anthropic, or Google key. The AI agent is a badge, a short thinking step, and a stamp. The count runs in your browser. Numbers stay on screen before any email ask. Email is only for the printable summary.

Is the AI agent calling a cloud model?

No. The AI agent on this page is the in-browser analysis experience. It uses rule-based heuristics. No squad design is sent to a third-party model.

How is this different from the squad sizing calculator?

Squad sizing estimates people and cost for a delivery plan. This page only pictures a supervisor-worker split and counts handoffs. Use both if you need a headcount and a picture.