When should I switch from one agent to multiple?

When a single agent's context fills up with role-switching prompts, or when separate specialists give you better tool-selection accuracy than one generalist. Both are the same symptom: the role boundaries got blurry.

Is supervisor always slower than handoffs?

Not always. Supervisor wins for known, well-bounded workflows where the orchestrator can integrate sub-agent outputs in parallel. It loses on long, exploratory tasks where the supervisor turns into a serial choke-point.

How do I debug a multi-agent system?

Per-step traces with the score deltas for tool selection at each step, plus an integration trace that shows the hand-off context. Without per-step observability, multi-agent debugging is guesswork.

Topic Pillar

Multi-Agent Systems.Supervisors, handoffs, swarms — and when each one breaks.

One agent works. Two agents talk. Five agents argue forever unless the orchestration is right. Multi-agent systems are not "more agents" — they are an architecture decision about who decides what, when, and with what budget. The orchestration pattern matters more than the model choice.

● 5 cluster pages· 2 posts· 3 updates

Supervisor vs handoffs

Supervisor pattern: one orchestrator delegates to specialists and integrates their answers. Clean on paper, slow in production when the supervisor becomes a bottleneck. Handoffs: agents pass the baton with the current context. Messier in code, recovers better when one agent loses the plot. Most teams that start with supervisor end up on handoffs for the recovery story.

When to reach for a swarm

Swarms shine when sub-goals are genuinely parallelisable and the work is exploratory — research, broad search, multi-source synthesis. Skip swarms for sequential workflows; the coordination overhead destroys the parallel-win on anything that has to happen in order.

Parallel sub-agent execution in practice

Modern agent runtimes can fan out to sub-agents for independent reads, then serialise on the write turn. This cuts a 20-minute refactor to under 4. The orchestration pattern matches the supervisor / swarm split — supervisor for the plan, swarm for the read phase, supervisor again for the write.

2 blog posts

Deep dives on Multi-Agent Systems

Multi-Agent

Why I am replacing supervisor patterns with handoffs

Supervisors looked clean on paper and shipped slow in production. Handoffs read messier in the code but recover better when an agent loses the plot. Two real systems and where supervisors still earn their keep.

Apr 26, 20266 min

Read the post Multi-Agent

Haiku 4.5 made our router 5x cheaper. The trade-off matters

Replacing Sonnet with Haiku in the dispatcher role cut our orchestration cost dramatically. It also cost us in two specific places I did not predict.

Feb 22, 20265 min

Read the post

3 ship-news updates

Latest in Multi-Agent Systems

Tools

Claude Code adds parallel sub-agent execution — multi-file refactors land in a single turn

May 13, 2026 · via Anthropic

Claude

Haiku 4.5 in production — small-model speed, surprising tool-use chops

Apr 22, 2026 · via Anthropic

Research

Anthropic research: when to use supervisor vs. swarm patterns in multi-agent systems

Apr 15, 2026 · via Anthropic Research

See the full Multi-Agent Systems update feed

Enterprise solutions

How Multi-Agent Systems ships in our engagements

The pages below are the buyer-focused, conversion-grade versions of this topic — deliverables, methodology, ROI, security considerations, and CTAs to scope a real engagement.

Solution