Vercel posted two agent changes on October 27.
First, Vercel Agent sessions in the dashboard and Slack now cost 67% less on average, by Vercel's count. The rates are the same: paid sessions bill at the model provider's inference rate plus a Vercel fee of $0.25 per million tokens. What changed is how many tokens a session burns. Vercel says it now manages context incrementally as conversations grow (long Slack threads included), keeps more reusable context in the prompt cache across turns, and ships updated CLI skills with clearer steps for common workflows and error recovery. Nothing to turn on.
Second, the Vercel Plugin works in OpenCode. It needs OpenCode 2, installs with one command, and adds skills for the AI SDK, Vercel Functions and Workflow, plus agents for deployment and performance work. /vercel:deploy deploys and /vercel:env manages environment variables. OpenCode 2.2.0 also came out the same day, with experimental desktop plugins and per-provider prompt cache TTLs.