Skip to main content

​
Key concepts

  • Tenant (company): the organization where agents run and tools are available.
  • Connector: stores application authentication and exposes tools.
  • Tool: an action an agent can take in a connected application.
  • Parent agent: the orchestrator that runs on chat or triggers.
  • Sub-agent: a specialized worker with prompts, tools, and memory.
  • Prompt template: a reusable prompt with variables and guardrails.
  • Trigger: a schedule or webhook that starts a workflow.
  • Rule: conditional logic that decides what runs next.
  • Approval: a user gate that must be approved before action.
  • Index data store: internal knowledge that agents can query.
  • Notifications: alerts about errors, approvals, and outcomes.
Diagram showing connectors feeding tools into parent agents and sub-agents with triggers, rules, and approvals

​
How users interact with agents

Chat is the main way users test agents and use them after setup. It gives users one place to ask questions, route work to parent agents, search internal data, and review workflow execution. Users can interact with agents in three ways:
  • Choose a default agent when every message should go to one parent agent.
  • Use Select Agent when Maadify should choose the best parent agent based on the question and each agent description.
  • Configure commands when users need quick access to specific agents from the command box.
Chat can also run Universal Search across selected index data stores. This helps users ask questions about internal information without choosing a workflow first. When a parent agent runs from chat, users can inspect the workflow in real time. Each step shows the agent response, tool call, or tool execution with the input and output. Responses can include sources from retrieved data, generated CSV output, or live action context such as a virtual browser session.

Chat with agents

Learn how chat routing, Select Agent, commands, workflow steps, and sources work.

​
How a workflow runs

1

Connect your tools

You connect applications and make tools available to your tenant.
2

Define agents and prompts

You create parent agents and sub-agents with prompt templates and memory context.
3

Add triggers and rules

You specify when agents run and what logic controls the next step.
4

Review and approve actions

Approvals and notifications keep humans in the loop before actions run.

​
Where to go next

Connectors

Configure authentication, tools, and defaults.

Parent agents

Build orchestrators with rules, approvals, and triggers.

Chat with agents

Test agents, route questions, and inspect workflow steps.

Agent activity

Investigate runs, tools, and context.

Index data store

Provide internal data as a tool for agents.