---
title: Recipes
description: Prompts you can hand to a coding agent to extend the eve Software Factory template, each grounded in the real files and APIs it needs to touch.
type: overview
summary: What a recipe is, how to run one safely, and the full list.
---

# Recipes



Each recipe is a prompt you copy into a coding agent. The prompt names the real files, APIs, and reference docs the change needs, so the agent edits the right things instead of inventing a plausible-looking approach.

Every recipe also tells you what it touches and what it costs you, so you can read it before you run it.

## The recipes

### How work arrives

<Cards>
  <Card href="/recipes/fire-on-every-issue" title="Fire on every issue" description="Run the pipeline on every new issue instead of waiting for the factory label." />

  <Card href="/recipes/run-on-a-schedule" title="Run on a schedule" description="Sweep for queued work on a cron instead of waiting for a webhook." />
</Cards>

### What the factory can do

<Cards>
  <Card href="/recipes/add-the-sentry-mcp" title="Add the Sentry MCP" description="Give the factory production error context through Sentry's hosted MCP server." />

  <Card href="/recipes/add-the-vercel-mcp" title="Add the Vercel MCP" description="Give the factory deployment and runtime context, app-scoped so unattended runs can use it." />

  <Card href="/recipes/request-your-review" title="Request your review" description="Put yourself on every pull request the factory opens, so it lands in your review queue." />

  <Card href="/recipes/merge-behind-approval" title="Merge behind approval" description="Let Foreman merge a pull request once a person clears the approval card." />
</Cards>

## How to run one

Every recipe edits source code that then has to be validated and redeployed, so start from a local clone of the [template repository](https://github.com/vercel-labs/eve-software-factory-template). There is nothing here you can change on a running deployment. When a recipe also needs a change in the target repository the factory works on, it says so under "What it touches."

<Steps>
  <Step>
    ### Read the recipe first

    Each recipe opens with its prompt. Everything below covers what the change does, which files it touches, what to weigh before running it, how to verify it, and common questions.
  </Step>

  <Step>
    ### Hand the prompt to your agent

    Copy the prompt into a coding agent that has the template checked out. Each prompt tells the agent to read the repository's own `AGENTS.md` and the relevant eve docs before writing code.
  </Step>

  <Step>
    ### Check the work before deploying

    Every prompt ends by telling the agent to run `pnpm validate`, which must report zero errors and zero warnings. Recipes that change routing or the tool surface also have it run `pnpm eval --tag fast`, and the ones that add a connection or a schedule have it run `npx eve info` so you can see the new capability in the discovered surface.

    Read what it reports rather than taking "done" at face value.

    No recipe deploys anything. Each prompt ends with "do not deploy", so shipping stays your call.
  </Step>
</Steps>

<Callout type="warn" title="Recipes name real files, so they age">
  Each prompt points at specific paths and identifiers in the template. If the template has moved on since a recipe was written, tell your agent to trust the repository over the prompt.
</Callout>

## Recipes and the docs

Recipes are for adding a capability the template does not ship with. The docs keep their own prompts for changes that belong next to the concept they modify: [deploying the template](/docs/getting-started), [changing the branch prefix](/docs/configuration), [keeping the evals in sync](/docs/evals), and [adding a fifth station](/docs/customization).

Before writing your own, [customizing](/docs/customization) covers where each behavior lives, and the [trust model](/docs/trust-model) covers the rules a new capability has to respect.


---

For a semantic overview of all documentation, see [/sitemap.md](/sitemap.md)

For an index of all available documentation, see [/llms.txt](/llms.txt)

For agent-facing discovery, including API and MCP surfaces, see [/agents.md](/agents.md)