FIG·05
THE DURABILITY CURVE INSTRUMENT · 05
THE MULTI-AGENT DECISION

One agent,
or many?

Most multi-agent systems are an org chart drawn in software: a researcher agent, a writer agent, a critic. That shape adds a coordination tax without adding capability, because the sub-tasks were never independent. Multi-agent earns its cost under one condition, and the number of agents is not it. List the jobs you are thinking of handing to a fleet, answer four questions on each, and this ranks them by whether multi-agent pays.

Runs entirely in your browser. Nothing leaves this device.
STEP 1 OF 3

What are you handing to a fleet?

Add the jobs you are considering splitting across agents, or one you already have. Pick the closest shapes or type your own. Judge the real task in front of you, not the demo version of it.

Common shapes
STEP 2 OF 3

Four questions each.

TASK 1
STEP 3 OF 3 · THE VERDICT

Your tasks, ranked by fit.

Ranked by whether multi-agent earns its coordination tax. Collapse means it is one job in a team costume. Tap any task to see its architecture, cost, and the anti-pattern to avoid.

Ranked by fit0
THE ARCHITECTURE
WHAT IT COSTS
THE ANTI-PATTERN

THE ONE-SENTENCE TEST

Which of these were you about to split, and can you say in one sentence why the split pays for itself?

THE DURABILITY CURVE
Instrument · 05 · Companion to “Your Multi-Agent System Is an Org Chart”