Configuration

Agentflow projects can be configured by creating an agentflow.config.js file in your project's root directory. This configuration defines the environment in which your workflows are executed, including AI providers, custom actions, tools, and plugins.

⚠️ Warning

Agentflow is a new project and the configuration API is highly likely to change as the project evolves. We recommend checking the documentation when upgrading to new versions for any breaking changes.

Basic Configuration

Create an agentflow.config.js file in your project root:

import { defineConfig } from '@agentflow/core'
import { openai } from '@ai-sdk/openai'
import { anthropic } from '@ai-sdk/anthropic'

import { debugAction, writeFileTool } from './lib/custom'

export default defineConfig({
  // Register custom actions
  actions: [
    debugAction
  ],

  // Register tools for LLM actions
  tools: [
    writeFileTool
  ],

  // Register LLM providers (compatible with Vercel AI SDK)
  providers: {
    openai,
    anthropic: createAnthropic({ apiKey: process.env.ANTHROPIC_API_KEY }),
  },

  // Add Agentflow plugins
  plugins: [],

  // Add custom workflow validators
  validators: []
})

Configuration Options

actions

Register custom actions for use in your workflows. Actions are components that work on input and generate a result. Agentflow includes several built-in actions (<GenText />, <GenObject />, <Loop />, and <Cond />), and you can create your own using the defineAction function.

tools

Register tools that can be used by LLM actions (<GenText /> and <GenObject />). Tools can be defined using the defineTool function, and any tools compatible with Vercel's AI SDK can also be used.

providers

Register AI providers that will be available to your workflows. Agentflow uses Vercel's AI SDK for provider compatibility, meaning you can use any provider that works with the AI SDK.

plugins

Register plugins that can extend the Agentflow environment. A plugin is a function that receives the Environment object and can add its own actions, tools, or providers.

validators

Add custom validation rules for workflow ASTs. These validators run when a workflow is compiled and can be used to enforce specific requirements or constraints.

Environment Variables

Agentflow will automatically load environment variables from a .env file in your project's root directory. This is particularly useful for storing sensitive information like API keys that shouldn't be committed to version control.

# .env
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-...

These variables are accessible via process.env in your configuration file, as shown in the example above where we configure the Anthropic provider.