· ai-devtools
Vercel Agent sessions cost 67% less due to context management improvements
Conversations within Vercel Agent, whether in the dashboard or Slack, now average out to 67% less in cost.
This reduction stems from a fundamental rebuild in how Agent handles context, leading to fewer tokens required per session and better utilization of prompt-cache reuse. These enhancements are active immediately and necessitate no user intervention.
Several specific improvements have been implemented. Updated Vercel CLI skills now furnish agents with clearer direction for typical workflows and when recovering from errors; other agents utilizing these skills also see the benefit.
Furthermore, context management is now more efficient as conversations deepen, successfully handling extensive Slack threads. Prompt caching has also seen an upgrade, allowing Agent to retain more reusable context across turns so that more information can be retrieved from the cache.
The guidance enhancements also translate to more efficient coding and planning, enabling Agent to draft code and map out changes using fewer tokens.
It is important to note that Vercel Agent’s core billing rates remain unaltered. Paid sessions are billed based on the model provider’s inference rates, supplemented by a Vercel Agent fee of $0.25 per million tokens.
You can track your usage and associated costs within the Agent section of your dashboard.
