Vercel AI Gateway adds a Sessions view for multi-turn agent costs
AI Gateway logs now group requests into sessions, showing cache hit rate, cost and which coding agent ran each conversation across every turn.
Vercel’s AI Gateway logs now have a Sessions view that groups requests into sessions and shows how session affinity routed each one, the company said on 17 October. Session affinity keeps a multi-turn conversation on the same provider so it can reuse cached context instead of reprocessing the prompt from scratch each turn.
Each session row shows the coding agent or app behind it, Claude Code, Codex or Cursor are named examples, plus its total cost, the models and providers used, and how much of its input was served from cache across turns. Opening a session shows its path: every turn with its model and timing. Sessions form automatically from a fingerprint of the conversation’s opening, or you can group them explicitly by sending the same sessionId in providerOptions.gateway or the same x-session-id header on every request. You can filter by turns, client, model, provider, authentication or cost, and single-turn sessions are hidden by default.
Why it matters: if you route a coding agent or chatbot through AI Gateway, this is the first place to look when a conversation gets expensive or slow, since it separates genuine cache misses from a provider simply not getting session affinity in the first place.
// group your own multi-turn requests into one session
headers: { 'x-session-id': conversationId }The caveat: session affinity and the grouping only apply to text and chat requests, and fingerprinted sessions are a best-effort grouping, not a guarantee, if your client never sends an explicit ID.