lg:agent

lg:agent - Agent Management

Manage agents in LangGraph projects including creating, modifying, and organizing multi-agent systems.

Workflow Steps

Add Agent

1. Validate Input

   • Check project structure exists
   • Verify agent name is unique
   • Validate agent type

2. Generate Agent Code

   • Create agent class/function
   • Setup LLM configuration
   • Add tool bindings
   • Configure system prompt

3. Update Graph

   • Add agent node to graph
   • Setup routing logic
   • Add conditional edges (if needed)
   • Update state schema (if needed)

4. Generate Tests

   • Create agent test file
   • Add unit tests
   • Add integration tests

5. Update Documentation

   • Add agent to README
   • Document agent capabilities
   • Update architecture diagram

Remove Agent

1. Analyze Dependencies

   • Find all references to agent
   • Check for dependent edges
   • Identify impacted routes

2. Remove Components

   • Remove agent node from graph
   • Remove agent file
   • Clean up edges
   • Update routing logic

3. Update Tests

   • Remove agent tests
   • Update integration tests

4. Update Documentation

   • Remove from README
   • Update architecture diagram

Modify Agent

1. Load Current Configuration

   • Read agent definition
   • Parse current settings

2. Apply Changes

   • Update specified parameters
   • Preserve unchanged settings
   • Validate new configuration

3. Update Tests

   • Adjust test cases
   • Add new test coverage

4. Update Documentation

   • Document changes
   • Update examples

Agent Types

ReAct Agent

Reasoning and Acting pattern with tool use.

from langchain_anthropic import ChatAnthropic
from langgraph.prebuilt import create_react_agent

llm = ChatAnthropic(model="claude-sonnet-5")
agent = create_react_agent(
    llm,
    tools=tools,
    state_modifier="You are a helpful assistant."
)

Tool-Calling Agent

Optimized for structured tool usage.