Learn
What's the difference between a master agent and specialist agents?
A master agent coordinates; specialist agents execute. The master (often a 'chief of staff' or 'orchestrator') delegates to specialist agents that each focus on one narrow domain or tool, then synthesizes their outputs. Specialists go deep on one thing; the master sees across all of them.
A master agent coordinates; specialist agents execute. The master (often a 'chief of staff' or 'orchestrator') delegates to specialist agents that each focus on one narrow domain or tool, then synthesizes their outputs. Specialists go deep on one thing; the master sees across all of them.
Splitting the work this way is what lets a system watch a whole stack at once. One giant model trying to do everything is hard to steer and easy to overwhelm; a master coordinating focused specialists is easier to reason about, cheaper to run, and simpler to extend — you add a capability by adding a specialist.
In Foundry, each specialist agent implements the same small interface and watches a single tool (a GitHub agent, a Vercel agent, a Stripe agent, and so on). Every specialist emits findings in one shared shape, which is exactly what lets the master synthesize across them without special-casing each source.
FAQ
- Why not use one big model instead of a master plus specialists?
- Focused specialists are easier to steer, cheaper, and independently improvable. The master's job — holding context across all of them and ranking what matters — is a different skill from any single specialist's.
See it working in Foundry.
Connect your stack and get one briefing a day.
Open Foundry →Free · built in public