Agentic System

Design and configure your multi-agent AI system with the visual agent builder.

The Agentic System tab is where you design your multi-agent system. The layout is a resizable two-panel view -- the left panel shows agent configuration, the right panel toggles between Chat and Deploy Logs.

How It Works

When you enter Control Hub, the AI Architect agent connects via WebSocket and generates a multi-agent architecture spec based on your use case, structured report, and app description. It may ask clarifying questions (MCQ or text) before generating the spec.

The spec arrives as a JSON config defining agents, each with: name, description, prompt, model, tools, memory, visibility, and agent-as-a-tool settings.

Agent Configuration

Select any agent card to expand it. You can view and edit:

  • Name & Description -- Agent identity and role
  • Model -- 24+ models available across providers:
    • Anthropic: Claude Opus 4, Claude Sonnet 4.5, Claude Haiku 4.5
    • Amazon Nova: Nova Pro, Nova Lite, Nova Micro, Nova Sonic (voice)
    • Meta Llama 3.3, Llama 3.2
    • Mistral Large, Mistral Small
    • NVIDIA Llama 3.1 Nemotron Ultra
    • Qwen 2.5 series, MiniMax, Z.AI GLM
  • Tools -- RAG, MCP, Web Search, Generate Report, Text to Speech, Speech to Text, Video tools
  • Memory -- Short-term (session), Long-term (persistent across sessions)
  • Visibility -- Both, User Only, Admin Only
  • System Prompt -- The full agent prompt, editable
  • Agent as a Tool -- Which other agents this agent can call

All editable inline, and changes can be saved or refined further via the chat panel.

Platform Agents

System agents that can be toggled on/off separately from your custom agents. These handle infrastructure, security, and observability concerns automatically.

Chat Panel

The right panel toggles between Chat and Deploy Logs:

Chat -- Refine your architecture via natural language. Send requests like:

  • "Add a sentiment analysis agent to the team"
  • "Change the coordinator to use long-term memory"
  • "Switch the research agent to Claude Opus 4"
  • "Enable voice on the support agent"

The agent responds with updated config or text explanation. If the spec is updated, the agent cards refresh on the left.

Deploy Logs -- Shows deployment progress with a real-time progress bar and job statuses.

Approve & Deploy

When satisfied with the configuration, click Approve. A Deploy Credentials dialog opens asking for your GitHub username and Bedrock API key. You can validate your Bedrock API key before proceeding.

The deployment provisions your agentic system on AWS Bedrock AgentCore. Deploy logs stream in real-time. On success, the Test Bed and Deploy tabs unlock.