Skip to main content
Bizzy puts AI to work in two forms: agents and automations. Both can read your data and act on it with tools — but an automation runs itself when an event fires, and an agent works in conversation with you.

​
Agents vs. automations at a glance

A useful intuition: an automation is a tireless junior employee waiting for the inbox to ping. An agent is a colleague you can ask things of.

​
What is an agent?

An agent is a configured persona. When you create one you decide:
  • Name and system prompt — who the agent is and how it behaves.
  • Model — which LLM powers it.
  • Tools — what it has access to, and at what permission level.
  • Step budget — how many tool calls a single reply can make. The default is 5.
  • Model parameters — temperature and other sampling settings, for power users.
Send the agent a message and it works step by step — reading, calling a tool, using the result to decide what to do next — until it’s done or it hits its step budget. Follow-up messages pick up right where the last one left off.

​
Resource permissions

An agent’s permissions control how it reads, writes, and deletes resources. Each action is allowed automatically, requires your approval, or is denied. Tools sharing a resource action share its setting. New agents start with the default agent permissions: reads and invoice draft writes are allowed; most other changes require approval. Children inherit their parent unless explicitly overridden. A resource with no permitted parent is denied. Saved settings stay in place when defaults change. Open the agent’s Tools tab to change resource permissions. Changes apply to the next chat turn, MCP request, or automation run. MCP and automations use only allowed actions because they cannot request chat approval.

​
Conversations

Every chat with an agent is recorded as an agent conversation — a thread you can revisit, share, or audit later. Each conversation has:
  • A status: active, completed, timeout, or error.
  • A source: web (the dashboard chat UI), api, or mcp (a Model Context Protocol client like Claude Desktop or Cursor).
  • The full transcript, including tool calls, tool results, and any approval decisions.
Browse the conversation history of any agent to see exactly what it did and why — review AI work without watching every step in real time.

​
How an agent runs

The loop is bounded — the step budget caps tool calls per reply — and interruptible: deny a permission prompt or pause the conversation and the agent stops cleanly.

​
When to use which

Reach for an agent when:
  • You want to ask follow-up questions in a conversation.
  • The work needs judgment, not just rules — “summarize this account’s recent activity,” “draft a polite refund refusal.”
  • You want a human in the loop on writes.
Reach for an automation when:
  • The trigger is an event you can describe — “a new email arrives,” “a customer signs up.”
  • You want it to run forever, untouched, until you tell it to stop.
  • The behavior is predictable enough that you don’t need to chat about it.
When in doubt: if you’d ever consider hitting “send” yourself before the action happens, you probably want an agent. If the action should always happen the moment the trigger fires, you probably want an automation.

Create an agent

Build your first agent in the web app

Resource permissions

Configure allow / ask / deny for each tool
Last modified on September 13, 2026