Core Concepts

This page explains the fundamental concepts you need to understand when working with Zenmako.

Agents

Agents are AI assistants that work on your behalf. Each agent has access to your connected tools and knowledge base, allowing it to understand your data and take meaningful actions.

When you ask an agent a question, it draws on indexed information from your connections to provide accurate, context-aware answers. Agents can also execute tasks like creating issues, updating records, or sending messages—depending on the permissions you have configured.

You can think of an agent as a team member who has read access to all your connected tools and can perform actions within the boundaries you set.

Connections

Connections are the links between Zenmako and your external business tools such as Asana, Trello, Notion, HubSpot, and others. Each connection stores the credentials needed to authenticate with the external service.

When you create a connection, you authorize Zenmako to access data from that tool. The connection defines what data can be read and what actions can be taken. You can have multiple connections to the same tool type if you work with different accounts or workspaces.

Connections are the foundation of everything else in Zenmako—without them, agents have no data to learn from and no tools to interact with.

Knowledge Base

The knowledge base is how Zenmako learns from your connected data. When you connect a tool, Zenmako indexes the relevant information—projects, tasks, documents, contacts, and more—so that agents can quickly retrieve and understand it.

This indexing process allows agents to answer questions about your data without needing to query external APIs in real time. For example, you can ask "What are the overdue tasks in Project Alpha?" and the agent will search the knowledge base for the answer.

The knowledge base stays synchronized with your connected tools, updating as your data changes. This ensures agents always have access to current information.

Workflows

Workflows are automated sequences of actions that run on a schedule. They allow you to set up recurring tasks or one-off jobs that execute without manual intervention.

A workflow might check for overdue tasks every morning and send a summary to Slack, or it might create weekly status reports from your project management data. Workflows combine the power of agents with scheduling, enabling hands-off automation.

You can configure workflows to run once at a specific time or on a repeating schedule (daily, weekly, etc.). Each workflow execution uses the same agent capabilities and respects the same permission boundaries as interactive requests.

Tool Permissions

Every connection has permission settings that control what actions agents and workflows can take. There are three permission levels:

  • Allowed: The agent can execute actions automatically without asking for confirmation. Use this for low-risk operations you trust completely.
  • Requires Approval: The agent will propose the action and wait for your explicit approval before executing. This is useful for operations that modify data or have meaningful consequences.
  • Denied: The agent cannot perform this action under any circumstances. Use this to block sensitive or destructive operations entirely.

    • Permissions give you fine-grained control over automation boundaries, ensuring agents only take actions you are comfortable with.