Features · Agent swarm

One request. Several agents working at once.

Some tasks aren't one question — they're several. Research a topic, write it up, chart the data, and draft an email about it. Simplicity's agent swarm splits work like that across multiple sub-agents running in parallel instead of making you wait through it one step at a time.

Four kinds of specialist agents

When a request is big enough to benefit from delegation, Simplicity spawns a team of up to three sub-agents, each assigned a kind — research for web-grounded fact-finding, writer for drafting documents, coder for building live app previews, or general for anything else. Each agent gets a clear, self-contained task rather than a vague slice of the whole request.

Watch it happen, not just wait for it

The swarm renders as a hub with each agent as a node radiating outward — the same visual language whether you're glancing at it in passing or want to dig into exactly what a research agent found. Hovering a node pops open its live steps and result summary right inline, without leaving the conversation.

Sub-agents can't send email or spawn more agents

Sub-agents are deliberately limited in scope — they can research, write, and build, but they can't send emails or spawn further sub-agents themselves. That keeps the blast radius of any single agent's output small: nothing leaves the conversation as an action (like a sent email) without a human reviewing and approving it first.

Frequently asked questions

What is the Simplicity agent swarm?

The agent swarm is Simplicity's way of handling large or multi-part requests. Instead of one model doing everything sequentially, Simplicity spawns up to three focused sub-agents — research, writer, coder, or general — that each take a piece of the task and work in parallel.

Can I see what each agent is doing?

Yes. Each agent appears as a node radiating out from a central hub, with a live status dot. Hover or click any node to see its current task, the tools it's using, and its progress — or open the full control room to watch every step across all agents at once.

What kinds of tasks use the agent swarm?

Anything that naturally splits into parallel workstreams: a research report that also needs a chart and a summary email, or a multi-part project like "research this topic, write it up, and build a simple demo page." Simplicity decides on its own whether a request needs one agent or a swarm.

Is the agent swarm slower or faster than a single AI response?

Faster for multi-part work, because the sub-agents run concurrently rather than one after another. A single simple question still gets a direct, immediate answer — the swarm only kicks in when a task actually benefits from being split up.