Building and Running Agents

What is an Orbitype AI Agent

A persistent, reusable specialist that runs autonomously on your workspace data. Give it a system prompt, a set of tools, and access to your connected resources. It reads context, makes decisions, calls tools, and acts until the task is done or it needs you. Build one agent per responsibility. Reuse it across workflows, artifacts, and other agents.

When do I use an agent instead of a workflow

When the task requires judgment, not just rules. Lead qualification, account research, outreach personalization, signal interpretation. If you cannot write a fixed rule for it, use an agent. If the logic is deterministic and repeatable, use a workflow. Most revenue motions need both.

What tools can an agent use

Any tool you give it. Same execution environment as workflows: AI (chat, image, embeddings), headless browser, web search, LinkedIn Advanced, OCR, XLSX, email, full database and S3 access. Write a tool function, describe it, attach it. The agent decides when to call it.

How does human-in-the-loop work

You define exactly where approval is required: before sending a message, before updating a deal stage, before escalating. The agent pauses, surfaces the decision to the right person, and waits. Everything else runs autonomously. Control what matters without becoming the bottleneck for everything else.

Can I import agents from the community

Yes. The Orbitype community ships reusable agent blueprints: ICP research, lead qualification, outreach personalization, signal monitoring, and more. Import directly into your workspace, configure with your data and credentials, deploy. No building from scratch required.