Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/vercel/eve/llms.txt

Use this file to discover all available pages before exploring further.

eve is a filesystem-first framework for building durable AI agents in TypeScript. You define what your agent can do by adding files to an agent/ directory — instructions, tools, skills, channels — and eve turns that structure into a running agent that persists sessions, serves HTTP, and connects to Slack, Discord, and other platforms.

Quickstart

Scaffold and run your first agent in under five minutes

Tutorial

End-to-end walkthrough: build an agent from scratch

Core Concepts

Understand the filesystem layout, tools, skills, and channels

TypeScript API

Every define* helper, runtime context, and import path

​
How eve works

An eve agent is a directory of files. Each file’s location says what it does: 
my-agent/
└── agent/
    ├── agent.ts            # model and runtime config
    ├── instructions.md     # always-on system prompt
    ├── tools/              # typed functions the model can call
    ├── skills/             # procedures loaded on demand
    ├── channels/           # HTTP, Slack, Discord, and more
    └── schedules/          # recurring cron jobs
When a message arrives — from a web UI, a Slack message, or a curl command — eve normalizes it, runs the model loop (including tool calls and subagent delegation), checkpoints progress at every step, and delivers the response back in the format the platform expects. Sessions are durable: they survive restarts, redeploys, and long pauses without losing context.

​
Key capabilities

Tools

Typed functions defined with defineTool and Zod schemas

Skills

On-demand markdown procedures the model loads when relevant

Channels

Connect to Slack, Discord, Teams, Telegram, Twilio, and more

Connections

Wire in external MCP servers and OpenAPI services

Subagents

Delegate focused subtasks to specialist child agents

Sandbox

Every agent gets an isolated bash environment automatically

Schedules

Run agents on a cron cadence with defineSchedule

Evals

Test agent behavior with defineEval and built-in reporters

​
Quick install

1

Scaffold a new agent

npx eve@latest init my-agent
Creates a new project directory, installs dependencies, initializes Git, and opens the interactive terminal UI.
2

Add a tool

agent/tools/get_weather.ts
import { defineTool } from "eve/tools";
import { z } from "zod";

export default defineTool({
  description: "Get the current weather for a city.",
  inputSchema: z.object({ city: z.string().min(1) }),
  async execute({ city }) {
    return { city, condition: "Sunny", temperatureF: 72 };
  },
});
3

Run the agent

npm run dev
The terminal UI opens. Type a message and watch the model loop run.
4

Deploy

eve deploy
Links to a Vercel project and ships the agent to production.
eve is currently in beta. The framework, APIs, and behavior may change before general availability. See the Vercel beta terms for details.

Build docs developers (and LLMs) love

Get started for freeTalk to us