Agent Graphs

Config Agent Graphs

You’re using a skill that will guide you through creating and managing agent graphs in LaunchDarkly. Your job is to design the graph topology, create it with the right edges and handoffs, and verify the routing between config nodes.

Prerequisites

This skill requires the remotely hosted LaunchDarkly MCP server to be configured in your environment.

Required MCP tools:

• create-agent-graph – create a new graph with nodes and edges
• get-agent-graph – inspect a graph’s structure and edges
• list-agent-graphs – browse existing graphs in the project

Optional MCP tools:

• update-agent-graph – modify edges, root config, or description
• delete-agent-graph – permanently remove a graph
• get-ai-config – inspect individual configs that serve as nodes
• create-ai-config – create new configs to use as graph nodes

Core Concepts

What Are Agent Graphs?

An agent graph is a directed graph where:

• Nodes are configs (each config is an agent with its own model, prompt, and tools)
• Edges define routing between configs (source -> target)
• Handoff data on edges controls how context is passed between agents
• Root config is the entry point — the first agent that receives user input

When to Use Agent Graphs

Graph Structure

[Root Config] --edge--> [Config A] --edge--> [Config C]
                  \--edge--> [Config B]

Each edge has:

• key – unique identifier for the edge
• sourceConfig – the config key that routes FROM
• targetConfig – the config key that routes TO
• handoff (optional) – data/instructions passed during the transition

Core Principles

1. Design Before Building: Map out nodes and edges on paper/whiteboard first
2. One Agent, One Job: Each node should have a clear, focused responsibility
3. Root Config Is the Router: The entry point should understand how to dispatch
4. Handoff Data Matters: Define what context flows between agents
5. Verify the Full Path: Test that routing works end-to-end

Workflow

Step 1: Design the Graph

Before creating anything:

1. Identify the agents (configs) needed — each is a graph node
2. Map the routing: which agent hands off to which?
3. Define handoff data: what context does each edge carry?
4. Identify the root config: which agent receives initial input?
5. Check existing graphs with list-agent-graphs to avoid duplicates
6. Check existing configs with get-ai-config to see what nodes already exist

Step 2: Ensure Nodes Exist

Each node in the graph must be an existing config. If configs don’t exist yet:

1. Use create-ai-config to create each agent config
2. Set up variations with appropriate models and prompts for each agent’s role
3. Verify each config exists with get-ai-config

Step 3: Create the Graph

Use create-agent-graph with:

• projectKey – the project containing the configs
• key – unique identifier for the graph
• name – human-readable display name
• description (optional) – explain the graph’s purpose
• rootConfigKey – the entry-point config key
• edges – array of connections between configs

{
  "projectKey": "my-project",
  "key": "support-triage-graph",
  "name": "Customer Support Triage",
  "description": "Routes customer queries to the appropriate specialist agent",
  "rootConfigKey": "triage-agent",
  "edges": [
    {
      "key": "triage-to-billing",
      "sourceConfig": "triage-agent",
      "targetConfig": "billing-specialist",
      "handoff": {"category": "billing", "priority": "normal"}
    },
    {
      "key": "triage-to-technical",
      "sourceConfig": "triage-agent",
      "targetConfig": "technical-specialist",
      "handoff": {"category": "technical", "priority": "normal"}
    }
  ]
}

Step 4: Verify

1. Use get-agent-graph to confirm the graph was created with the correct structure
2. Verify edges connect the right source and target configs
3. Check that the root config key matches the intended entry point
4. Confirm handoff data is present on edges that need it

Report results:

• Graph created with N nodes and M edges
• Root config set correctly
• All edges verified

Edge Cases

What NOT to Do

• Don’t create a graph before the config nodes exist
• Don’t forget handoff data when agents need context from predecessors
• Don’t create overly complex graphs — start simple and add nodes as needed
• Don’t delete a graph without understanding if it’s actively used in agent workflows

Other Resources

To learn more, read Agent graphs (opens in a new tab).