Ensure each node's output matches the input requirements of connected nodes.
How to/Evaluate Chains
Build complex AI workflows with Prompt Chains
Connect prompts, code, and APIs to create sophisticated AI systems using our visual editor - no coding required
When building AI applications, you often need multiple steps working together - like chaining multiple AI models, processing data, or calling external services. While single Prompts work well for simple tasks, complex workflows need a more structured approach.
What are Prompt Chains?
Prompt Chains let you connect three types of nodes to build your AI workflow:
- Prompt: Run LLM interactions
- Code: Add custom logic
- API: Connect external services via HTTP API
Each node's output feeds into the next, creating a seamless flow of data.
Available nodes
Create your first chain
Initialize your chain
Start from the Start node by dragging to create your first connection.
Choose node functionality
Configure the type and settings for your new node based on your requirements.
Expand the workflow
Connect additional nodes by dragging from the output handles to create new steps.
Version and deploy
Create versions to test or deploy your chain:
- Click
Savein the top right - Add an optional description
- Click
Update
View all versions using the icon
Next steps:
Learn more about testing prompts or check our SDK documentation.
Set up a human annotation pipeline
Human annotation is critical to improve your AI quality. Getting human raters to provide feedback on various dimensions can help measure the present status and be used to improve the system over time. Maxim's human-in-the-loop pipeline allows team members as well as external raters like subject matter experts to annotate AI outputs.
Debug AI agent errors step by step
Identify and fix errors at each step of your AI workflow with detailed diagnostics