agentic.md Build AI Agents & Agentic Workflows in Markdown

Define, visualize, and execute AI Agents and Agentic Workflows using Markdown and Mermaid diagrams. Seamlessly integrates with the Drivly AI ecosystem.

Define an Agent in Markdown

# Customer Support Agent

## Description
Handles customer inquiries and resolves common issues

## Properties
- name: Amy
- role: Customer Support Agent
- integrations: chat, slack, email, zendesk, shopify

## Workflow

```mermaid
stateDiagram-v2
    [*] --> Idle
    Idle --> ProcessingInquiry: onMessageReceived
    ProcessingInquiry --> ResolvingIssue: issueIdentified
    ProcessingInquiry --> EscalatingIssue: complexIssueDetected
    ResolvingIssue --> Idle: issueResolved
    EscalatingIssue --> Idle: escalationCompleted

Generate and Use an Agent

import { createAgentFromMarkdown } from 'agentic.md'
import fs from 'fs'

// Load markdown file
const markdown = fs.readFileSync('./customer-support-agent.md', 'utf-8')

// Create agent from markdown
const customerSupportAgent = createAgentFromMarkdown(markdown)

// Use the agent
customerSupportAgent.onMessageReceived({
  from: '[email protected]',
  content: 'I need help with my recent order'
})

Generate XState Machines from Mermaid Diagrams

import { createWorkflowFromMarkdown } from 'agentic.md'
import { interpret } from 'xstate'
import fs from 'fs'

// Load markdown with Mermaid state diagram
const markdown = fs.readFileSync('./support-workflow.md', 'utf-8')

// Generate XState machine from Mermaid diagram
const machine = createWorkflowFromMarkdown(markdown)

// The generated machine structure:
// {
//   id: 'workflow',
//   initial: 'Idle',
//   states: {
//     Idle: {
//       on: { onMessageReceived: 'ProcessingInquiry' }
//     },
//     ProcessingInquiry: {
//       on: {
//         issueIdentified: 'ResolvingIssue',
//         complexIssueDetected: 'EscalatingIssue'
//       }
//     },
//     ResolvingIssue: {
//       on: { issueResolved: 'Idle' }
//     },
//     EscalatingIssue: {
//       on: { escalationCompleted: 'Idle' }
//     }
//   }
// }

// Use the machine with XState
const service = interpret(machine)
  .onTransition(state => console.log(state.value))
  .start()

// Send events to the state machine
service.send('onMessageReceived')

Define a Workflow in Markdown

# Order Processing Workflow

## Description
Handles the end-to-end order processing flow

## Workflow

```mermaid
stateDiagram-v2
    [*] --> Idle
    Idle --> Processing: ORDER_RECEIVED
    Processing --> Completed: COMPLETED
    Processing --> Failed: FAILED
    Completed --> [*]
    Failed --> [*]

Compose Multiple Workflows

import { createWorkflowFromMarkdown, composeWorkflows } from 'agentic.md'
import fs from 'fs'

// Load multiple workflow definitions
const orderWorkflow = createWorkflowFromMarkdown(
  fs.readFileSync('./order-workflow.md', 'utf-8')
)

const paymentWorkflow = createWorkflowFromMarkdown(
  fs.readFileSync('./payment-workflow.md', 'utf-8')
)

// Compose workflows into a single state machine
const composedWorkflow = composeWorkflows({
  orders: orderWorkflow,
  payments: paymentWorkflow
})

// Use the composed workflow
const service = interpret(composedWorkflow).start()

Integration with Drivly AI Ecosystem

import { createAgentFromMarkdown } from 'agentic.md'
import { AI } from 'ai'

// Create agent from markdown
const agent = createAgentFromMarkdown('./agent.md')

// Use with AI
const workflow = AI({
  processCustomerInquiry: async ({ ai, event }) => {
    const response = await agent.processInquiry(event.message)
    return response
  }
})

Features

• Markdown-First Development: Define agents and workflows in plain Markdown
• Mermaid → XState: Generate state machines directly from Mermaid diagrams
• AI Agent Generation: Create autonomous agents from Markdown definitions
• Composable Architecture: Build complex systems from simple components
• TypeScript Support: Full type safety and autocompletion