Skip to main content

What is AgentMap?

Build Multi-Agent AI Workflows with Simple CSV Files

AgentMap is a framework for creating sophisticated AI workflows where multiple agents collaborate to solve complex problems. Instead of complex YAML or code, you define your workflows using familiar CSV files.

When Should You Use AgentMap?​

AgentMap is perfect when you need:

✅ Multi-step AI processing - Chain together different AI operations
✅ Version-controlled workflows - CSV files work great with Git
✅ Rapid prototyping - Build and modify workflows quickly
✅ Production orchestration - Scale from prototype to production
✅ Mixed AI providers - Combine OpenAI, Anthropic, Google in one workflow

Core Concepts (3 Minutes)​

🤖 Agents​

Agents are the workers in your workflow. Each agent has one job:

class WeatherAgent(BaseAgent):
    def process(self, inputs: Dict[str, Any]) -> Any:
        location = inputs.get('location', 'Unknown')
        # Call weather API and return data
        return f"Weather in {location}: Sunny, 72°F"

Built-in Agent Types: LLM agents, file operations, data processing, routing, user input.

📋 Workflows (CSV Files)​

Your workflow is a CSV file that defines how agents connect:

graph_name,node_name,agent_type,next_node,input_fields,output_field,prompt
HelloWorld,Start,input,Greet,,name,"What's your name?"
HelloWorld,Greet,echo,End,name,greeting,"Hello {name}!"
HelloWorld,End,echo,,greeting,result,"Thanks for using AgentMap!"

Key Elements: Nodes (agents), connections (next_node), data flow (input_fields → output_field).

🔧 Services​

Services provide infrastructure like LLM APIs, databases, and file storage:

# Agents automatically get the services they need
class MyLLMAgent(BaseAgent, LLMCapableAgent):
    def process(self, inputs: Dict[str, Any]) -> Any:
        # self.llm_service is automatically injected
        response = self.llm_service.call_llm(
            provider="openai",
            messages=[{"role": "user", "content": inputs["question"]}]
        )
        return response["content"]

Simple Example: Document Analysis​

Here's a complete 3-agent workflow:

graph_name,node_name,agent_type,next_node,input_fields,output_field,prompt
DocAnalyzer,LoadDoc,file_reader,Analyze,file_path,document,
DocAnalyzer,Analyze,llm,Summarize,document,insights,"Extract key insights: {document}"
DocAnalyzer,Summarize,llm,End,insights,summary,"Create summary: {insights}"
DocAnalyzer,End,echo,,summary,result,"{summary}"

What happens:

  1. LoadDoc reads a document file
  2. Analyze extracts insights using AI
  3. Summarize creates an executive summary
  4. Data flows automatically between agents

Ready to Build?​

Next Step: Quick Start →
Install and run your first workflow in 5 minutes

Alternative Learning Paths: