eve logo

eve

The Framework for Building Agents

Website GitHub

What is it?

What it is

A filesystem-first framework for building durable AI agents, where agent capabilities such as instructions, tools, skills, channels, and schedules are organized in conventional directories under an `agent/` folder.

Why it exists

To simplify the inspection, extension, and operation of AI agent projects by placing core agent logic in predictable filesystem locations, reducing configuration overhead and providing a clear authoring interface.

Who should use it

Developers familiar with TypeScript/JavaScript who want to build durable AI agents with a clear, filesystem-based structure; teams looking for an inspectable, extensible agent framework.

Who should avoid it

Those new to TypeScript or AI agent concepts, users seeking no-code/low-code solutions, or anyone requiring a stable, production‑grade framework (eve is currently in beta).

How it works

A quick walkthrough in plain English

How eve works in 4 steps

Step 1 of 4

Something triggers the flow

A schedule, webhook, or manual click tells eve to start.

Features

Agent
Framework
Harness
Javascript
Markdown
Sandbox

Advantages

  • Open source (Apache-2.0)
  • Active TypeScript ecosystem
  • Self-hosted deployment options

Disadvantages

  • Requires operational ownership for self-hosted setups
  • Community support varies by project maturity

Installation

native

Read the [documentation](https://eve.dev/docs) for the full project layout and guides.

## Quick start

FAQ

How do I create a new eve agent project?

Run `npx eve@latest init my-agent` to scaffold a new agent directory, install dependencies, initialize Git, and start the interactive terminal UI. To add eve to an existing project, navigate to the project and run `npx eve@latest init .`.

What files are required in an eve agent's `agent/` directory?

The only required file is `agent/instructions.md`, which contains the always-on system prompt. Other files like `agent.ts`, tools, skills, channels, and schedules are optional.

How do I define a tool that the agent can call?

Create a TypeScript file under `agent/tools/` (e.g., `get_weather.ts`). Import `defineTool` from `eve/tools`, define a description, input schema using Zod, and an `execute` function that returns the tool's result.

How do I configure the language model for my agent?

Edit (or create) `agent/agent.ts` and export a default agent via `defineAgent`. Set the `model` field to the desired model identifier, for example `model: "anthropic/claude-sonnet-5"`.

How do I start developing and testing my agent locally?

After scaffolding or configuring your agent, run `npm run dev` (or `npx eve dev`) to start the agent in development mode. This launches the interactive terminal UI where you can interact with the agent.

Where can I find the full documentation and tutorials for eve?

The documentation is bundled with the package and can be read locally from `node_modules/eve/docs`. Online, visit https://eve.dev/docs for guides, the first-agent tutorial, and API reference.

Loading documentation…
View on GitHub

Featured in Videos

YouTube tutorials and walkthroughs for eve

Alternatives

Similar projects ranked by category, topics, and text overlap.

Compare
eve | MushyBook