Projects pay the bills. Retainers build the studio. Four shapes — Care, Growth, Agent Ops and Partner — month-to-month, cancel anytime. Agent Ops is the newest and the most defensible: keeping live agents tuned as foundation models drift, token costs shift and guardrails need updating.
Each retainer comes with a defined scope, a monthly hour bank, an on-call commitment and a written SLA. No "ask us for anything" ambiguity — that's how retainers quietly kill agencies.
This is the scope that justifies the price. If a provider is charging you for "AI support" without doing most of these, they're selling a logo, not operations.
Usually, yes — especially after an agent ships. Marketing sites can survive without one if nothing ever changes. Live agents and SaaS products essentially can't; something drifts every month even if your code doesn't change, because the world around it does.
Up or down, any month, no penalty. The usual path is Care → Growth → Agent Ops as complexity grows, and the usual reason to downgrade is that a feature you launched stabilized — which is a good problem to have.
We'll flag it before we bill it. Overage is charged at the retainer's hourly equivalent — no surprise 2× rate. If overages happen two months in a row, we'll recommend bumping to the next tier.
Agent Ops can, usually. We run a short (usually one-week) audit first to make sure the agent is something we can responsibly operate. If the original build is structurally unsafe, we'll say so and scope a rebuild.
Partner tier comes close — 4h response, on-call for production incidents. For hard 99.9% uptime SLAs with penalties, we'll work with you on a custom contract. That's not a drop-in retainer, it's a real agreement.
If you've shipped with us, a retainer is the usual next step. If you shipped with someone else, we can audit and onboard — book a 30-minute call and we'll scope it.