Architecture

AgentMark uses a multi-stage transformation pipeline to convert your .prompt.mdx files into model-specific formats while maintaining a unified interface.

​

Processing Pipeline

​

1. MDX Processing

Your .prompt.mdx file contains:

• Frontmatter configuration
• Message components (<System>, <User>, <Assistant>)
• Dynamic content (props, components)

​

2. AST Generation

The Loader class has a method called load that parses your MDX into an Abstract Syntax Tree (AST), which:

• Validates syntax
• Processes imports
• Provides an AST that the SDK can use to generate prompts

​

3. Transformations

Two transformations occur:

​

3a. AgentMark Format

Agentmark client compiles the AST into default agentmark format.

• Standardized message structure
• Normalized model configuration
• Evaluates props

​

3b. Adapter translation

Adapter converts AgentMark format into adapter-specific formats:

• Message formatting
• Parameter mapping
• Tool/function calling adaptations

​

Implementation Example with Vercel AI SDK adapter

import {
  VercelAIModelRegistry,
  createAgentMarkClient
} from "@agentmark/vercel-ai-v4-adapter";
import { FileLoader } from "@agentmark/agentmark-core";
// use any ai-sdk model provider
import { openai } from "@ai-sdk/openai";

const modelRegistry = new VercelAIModelRegistry();
// register models you want to use
modelRegistry.registerModels(['gpt-4o'], (name: string) => {
  return openai(name);
});

const loader = new FileLoader("./path/to/agentmark");

const agentmark = createAgentMarkClient({
  loader,
  modelRegistry,
});

const run = async () => {
  const examplePrompt = await agentmark.loadTextPrompt("<example>.prompt.mdx");
  const props = { name: "Emily" };
  const vercelInput = await examplePrompt.format({props});
}
run();

​

Key Benefits

1. Abstraction: Developers work with a single, unified format
2. Portability: Prompts can be easily switched between models
3. Extensibility: New model support via plugin system
4. Type Safety: Full type-safety support throughout pipeline
5. Tooling: We can support a rich ecosystem of development tools