Write it once.
Your agents get it.
Briefs are the source of truth for what you want agents to do. Plain English, typed outputs, linked knowledge. One brief, many runs.
Three parts. No DSL.
Prose
Describe the job in plain English. Markdown, mentions, inline variables. The brief is the spec — no need to restate it in code.
Typed outputs
Define the shape of the answer once. Every agent subscribed to the brief produces the same structured output — ready for your code.
Linked knowledge
Attach docs, SOPs, URLs. Grain retrieves them at runtime. Update the doc → every agent picks it up on the next run.
Change the brief.
Every agent listens.
A brief isn't attached to an agent — agents subscribe to it. One brief can drive your prod deployment, staging twin, ten eval agents, and the human on-call. Update the source, everyone moves.
Briefs in the wild.
The multi-agent command center. Hire an entire AI team in 30 seconds.