Fan out to eve subagents with structured results

An eve workflow tool, review_pack, that sends each specialist its documents in a self-contained brief, runs the two declared subagents in parallel with ctx.agent and a Zod outputSchema, and merges their typed findings in code, flagging contradictions between them. It runs on OpenRouter, a local Ollama model, or scripted mock models for a free end-to-end check.

Back to the exercise

Before you start

Before you start:Node.js, pnpm and TypeScript Durable agents with eve Get an OpenRouter key Local models with Ollama

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? →

The lead here is not a model deciding whom to call; it is a workflow tool. review_pack reads the pack, writes each specialist a complete brief, calls both with ctx.agent(name).send(brief, { outputSchema }) in parallel, and merges what comes back in plain TypeScript. The root model only decides to call review_pack and then writes up its result. The decision that matters most is that the subagents are hidden from the root model (tool: false), so the only way to reach them is through the tool that enforces the schema.

Answers Lesson 1: Multi-agent systems

Step-by-step answer

Stretch 90 min8 steps