AgentMark takes a developer-first approach to prompt management, treating prompts as files that live in your repository while still providing a platform for non-technical team members. Prompt Management

​
Version-Controlled Prompts

All prompts in AgentMark are stored directly in your git repository as AgentMark files (.prompt.mdx extension). This approach provides several benefits:
  • Version Control: Track changes, manage branches, and review prompt updates like code
  • Local Development: Work on prompts using your preferred development tools
  • CI/CD Integration: Include prompt testing in your deployment pipeline
  • Collaboration: Use familiar git workflows for prompt management
  • AI-enhanced tooling: Use AI to generate prompts, test them, and even generate code to implement them

​
AgentMark Format

Prompts are written in AgentMark format, which combines markdown-like syntax with powerful templating and configuration options. Learn more in our AgentMark documentation.
example.prompt.mdx
---
name: example
text_config:
  model_name: gpt-4o-mini
---

<User>Hello World</User>
Name should be unique across all prompts in the repository. Also name should match the file name.

​
Multiple Ways to Edit

AgentMark provides multiple ways to work with prompts, accommodating different team members and workflows:

​
For Non-Technical Users

  • Web Platform: Edit and test prompts through our intuitive web interface
  • Instant Publishing: Changes are automatically committed to your repository
  • Integrated Testing: Test prompts and view results directly in our platform

​
For Developers

  • VSCode Extension: Test prompts easily with our VSCode Extension
  • Local Development: Use our CLI and SDK for local testing and development
  • Git Workflow: Work with prompts using your normal development workflow

​
Folder Structure

Prompts are typically organized in your repository like this: 
your-repo/
  agentmark/
    example1.prompt.mdx
    example2.prompt.mdx

​
Generation Types

AgentMark supports multiple generation types for prompts, including Text, Object, and Image. Each type is suited for different use cases, such as natural language, structured data, or visual content. Read more about Generation Types

​
Local Development

Visit our Development page to learn how to test prompts locally.

​
Have Questions?

We’re here to help! Choose the best way to reach us: