Deploy to Vercel with a deployable model

The eve agent made deployable to a personal Vercel Hobby project, with the OpenRouter key as a sensitive project variable, route auth that accepts only Vercel OIDC callers, token-based session limits, redacted Agent Runs export, and smoke checks for the health route and eve's own 401. Rehearsed locally on a production build first, then the exact deploy commands, and a list of every place client data would flow.

Back to the exercise

Before you start

Before you start:Install Claude Code Node.js, pnpm and TypeScript Durable agents with eve Get an OpenRouter key Deploy to Vercel

This exercise calls a model. Run it either way

Any model through OpenRouterOne key for Claude, GPT, Gemini, Llama and more. Pay per token, with a credit limit.A local open-source modelFree and private with Ollama. Weaker on hard tasks.

Which should I pick? →

Deploying is a short list of commands; the work is in what the deployment is allowed to do. The model must be one a deployment can reach and pay for: a local Ollama model can't be reached from Vercel, and eve refuses chatgpt() outright, so this agent uses an OpenRouter model instance whose key lives only in the Vercel project's environment. The decision that matters most is the spending guard. eve's costUsd is only filled in for models it has a price for, so for an OpenRouter instance the sessionUsage limit has to count tokens, and the real cap in pounds is the credit limit on your OpenRouter key.

Answers Lesson 6: Shipping agents

Step-by-step answer

Stretch 45–60 min8 steps