When to use ZDR routing instead of direct model calls?
Teams that must guarantee no provider retains their prompts or code are stuck when a model is only served by one provider without a ZDR agreement. Vercel AI Gateway now routes Kimi K3.1 Pro and its Highspeed variant across multiple ZDR providers, failing requests rather than falling back to non-ZDR ones.
What it is
AI Gateway is a unified API for calling models with usage tracking, retries, failover, and Zero Data Retention support. Kimi K3.1 Pro now has multiple ZDR providers, with a 1M-token context window, image input, and strong coding and agent-run performance.
Applied case
A developer running Claude Code installs the latest Vercel CLI, runs setup, and selects moonshotai/kimi-k3.1-pro or moonshotai/kimi-k3.1-pro-highspeed in the agent. They enable ZDR in the AI Gateway dashboard so every team request, including agent traffic, only routes to providers holding a ZDR agreement. For interactive coding they pick Highspeed, where response time matters; for long agent runs from a small fix to a full project they use the standard model.
- Enable ZDR once in the dashboard to enforce it for all team requests, coding agents included
When to use it
Use when compliance requires zero data retention, or you want failover across providers for higher-than-provider uptime.
Skip when a single direct provider call is fine and you have no retention constraints.
Read more
Recommended for you
- WorkflowOpencode
How do you keep coding agents on task without babysitting every step?
Agent sessions drift from the objective, forcing you to watch every turn; opencode's /goal lets a session run until a reviewer confirms the goal is met.
- WorkflowHerdr
How do you supervise coding agents without watching every window?
I run Claude and opencode side by side in one terminal. Three herdr habits replaced window-polling with state-waiting: persistent sessions, the agent loop, a setup script.
- WorkflowSuperpowersClaude Code
Why does your multi-agent workflow keep colliding?
Two agents in two threads share files but not context. Both decide on stale state. Fix: one fresh agent per task, with isolated context.
- WorkflowSuperpowersClaude Code
What makes agent output trustworthy when the model is not the problem?
The reliability problem is not the model. It is the missing disciplines around it. Brainstorm, plan, test, verify, review: each gate before the next step.
Enjoyed this article?
Subscribe for new articles. No spam. Unsubscribe anytime.