Apps

Apps in Hypermode are how you package and deploy agents and AI-native apps.

An App is a collection of related agents, tools, and memory, working together to perform a cohesive set of tasks towards an outcome. Whether you’re coordinating a team of agents or creating multi-step agentic flows, Apps are the top-level construct that unifies your AI system.

Why Apps?

Building an agent is powerful. But real-world use cases rarely stop at one. As your AI system grows, you’ll often need to:

• Coordinate multiple agents with different roles
• Share tools and context across flows
• Persist memory and long-term learning for a domain
• Deploy and version your system as a single unit

That’s where Apps come in. Apps give structure to your agentic architecture. They let you group related components - agents, tools, memories, and APIs- into one deployable unit.

What’s in an App?

Apps in Hypermode are modular and flexible, designed to let you build complex systems by assembling reusable components.

Every App on Hypermode is made up of the following components:

Agents

Agents are the workers that reason, plan, and act. Apps can include one or many agents, each with their own role. You can assign different models, tools, and memory configurations to each agent, or allow them to collaborate via shared context.

Tools

Tools are how agents take action. These include custom functions, external APIs, or built-in Hypermode tools (like data fetching or search). Apps define which tools are available to which agents and can scope tools to specific tasks or agent roles.

Memory

Apps can define long-term and short-term memory for agents using Hypermode’s memory primitives. This allows your agents to remember past interactions, user preferences, task outcomes, and more to enable persistent, contextual behavior over time.

Connections

Apps can include third-party integrations like GitHub, Slack, or internal APIs via the Model Context Protocol (MCP). These integrations allow agents to interact with external systems in a structured, secure way.

Decision interface

Apps use a decision interface to let you work asynchronously with agents. This is useful for long-running tasks, approvals, or cases where agent actions need to be reviewed before execution.

Runtime

Each App includes metadata for tracking versions, environments, and ownership. You can define environment variables, set up deployment environment (staging vs production), and manage runtime settings that affect how agents are executed.

Agent development lifecycle

Apps support the full development lifecycle:

• Develop locally using the Hyp CLI or in a conversational interface using Threads
• Test and debug agent behavior
• Deploy to production with versioning and rollback support
• Monitor and trace interactions using built-in observability
• Collaborate and share across teams with access controls and roles