Vercel Agent sessions now cost 67% less

Vercel said on 27 October that Agent sessions in the dashboard and in Slack now cost 67% less on average, with no action needed and no change to the billing rate itself.

The saving comes from how Agent manages context, not from a price cut. Vercel rebuilt its CLI skills to give Agent clearer guidance for common workflows and error recovery, made context tracking more efficient as a conversation grows (including long Slack threads), and improved prompt-cache reuse across turns. Together those cuts reduce the tokens a session burns through before it gets to an answer. The fee itself stays at $0.25 per million tokens on top of the model provider’s own inference rate.

Why it matters: if you or your team lean on Vercel Agent for day-to-day questions in Slack or the dashboard, the same usage should now show up noticeably cheaper on your bill without you touching a setting. It’s also a reminder that a chunk of what you pay an agent for is the context it drags around, not just the model call, so efficiency work on that side can move the bill as much as a price cut would.