Skip to main content
A Hop is an AI agent that works in your team’s Spaces. Not a general chatbot, not a do-everything AI. It’s an agent with a specific job, working from the context your team has built up.

How agents work in Hops

Every Hops workspace comes with a default agent called Hops. It’s available in every Space, handles general tasks, and responds when you tag @Hops in a thread. Beyond the default agent, you can create custom agents for any role your team needs. A marketing agent for copy and campaigns. A sales agent for account prep and follow-ups. A finance agent for reports and summaries. A support agent for tickets and escalations. Whatever fits your workflow. A Space can have multiple agents working side by side. Tag any of them with @ in the same conversation, and each one responds from its own specialization and context.

How connections work

The default Hops agent and your personal agent both use your workspace’s default connections. Any tool connected at the workspace level is available to them automatically. Custom agents can have their own connections on top of that. You choose which apps each custom agent connects to, so each one only accesses the tools relevant to its role.

What makes each agent unique

Every agent can be customized with:
  • Instructions that shape how the agent behaves, what it prioritizes, and what it avoids
  • Artifacts like documents, tables, and other reference material that give the agent specialized context
  • Custom connections giving per-agent access to specific apps and data from 500+ integrations across productivity, developer tools, CRM, communication, analytics, design, databases, and more

What makes agents in Hops different

They know what the team knows. An agent reads from your team’s actual context: threads, docs, decisions, and connected tools. You don’t re-explain the project every session. They show their work. Every agent response includes what it used and what it did. Nothing happens in a black box. They hand off cleanly. Work moves between agents and teammates with context attached. One agent can pick up where another left off without anyone re-briefing. They wait for your approval on risky actions. Before an agent sends something external, updates a record, or takes an action with real-world effects, it stops and asks you.

Agents are not

  • A replacement for your team
  • An autonomous system that acts without you
  • A general-purpose chatbot with no specialization
  • A black box that “just figures it out”
The human stays in the frame. An agent moves work forward. You make the decisions.

Types of agents