What is an AI Agent
An xpander AI Agent is designed to autonomously perform tasks by dynamically navigating through a graph of actions. Each node in the graph represents a specific function or step in the workflow, and as the AI Agent progresses through the graph, it exposes new functions or operations based on the current task context.
This graph-based approach enables AI Agents to handle complex, multi-step processes where each action is determined by the output of the previous steps. As the agent moves further along the graph, it can branch into Topics (or sub-graphs), where additional logic or steps are required. These Topics act as specialized segments, allowing the agent to handle more intricate tasks or workflows in a structured yet dynamic way.
Topics are enforced as the agent transitions into sub-agent graphs, ensuring that specific actions are executed within a controlled context. This mechanism allows the AI model to make intelligent decisions at each step while maintaining consistency and coherence across the workflow, adapting as needed to the evolving task or data.
Agent Builder
The AI Agent Builder allows developers to design and deploy AI Agents that can autonomously perform tasks. It provides a visual and intuitive interface, helping developers define workflows where each step and parameter is dynamically selected by the AI model. This flexibility enables the rapid development of intelligent agents without needing to predefine every possible workflow or scenario.
Agentic Interfaces
Agentic Interfaces are pre-built or custom integration points that allow AI Agents to interact with external systems. These connectors enrich the Agent’s capabilities by providing access to APIs, data sources, and services. The AI model dynamically retrieves and uses the required data or services from these interfaces , selecting appropriate parameters based on task requirements. With Agentic Interfaces, xpander agents can seamlessly integrate into a wide variety of enterprise applications and systems.
Agent Graph System
The Agent Graph System is the core mechanism that defines how AI Agents execute tasks. Instead of following static workflows, the system allows the AI Agent to dynamically choose each step in the task flow. The graph represents tasks as nodes (actions, API calls), and the AI Agent navigates this graph in real-time, selecting the next step and required parameters based on current data and conditions. This system ensures flexibility, as the agent can adapt the workflow based on the evolving needs of the task, handling complex, branching scenarios with ease.
Topics
Topics are used to enforce specific actions or behaviors within a the Agent Graph System. They ensure that the AI Agent only performs certain actions or uses certain data within the defined context. By grouping related Operations and Functions together under a topic, the AI Agent maintains a consistent approach to specific parts of the workflow, while still adapting dynamically. Topics are essential in keeping the agent’s decision-making structured and focused.
Sources and triggers
Sources represent the entry points for data or actions within the Agent Graph System. Each AI Agent can have multiple Sources, which serve as the initial triggers for the agent’s workflow. For example, an AI Agent might receive messages from Slack, process commands from Microsoft Teams, handle incoming data from a Webhook, interact with users through a WebUI, or be invoked via an SDK. These diverse entry points allow the AI Agent to integrate seamlessly with various platforms and services, ensuring it can respond to a wide range of inputs and scenarios.
Types of AI Agents
Serverless AI Agents
Serverless AI Agents in xpander.ai provide a fully managed solution for building intelligent agents, such as Slackbots or Microsoft Teams integrations, without the need to manage the underlying runtime or LLM integration. By leveraging the xpander-SDK and the Agent Graph System, these agents can autonomously perform tasks and interact with various platforms. The system optimizes prompts and handles LLM interactions automatically, eliminating the need for manual prompt engineering or orchestration. This makes Serverless AI Agents ideal for quickly deploying AI-powered workflows that scale effortlessly, all while xpander manages the backend infrastructure and AI logic.
Custom AI Agents
Custom AI Agents in xpander.ai offer enterprises the flexibility to build AI Agents using their preferred AI frameworks, such as NVIDIA NeMo, Amazon Bedrock, or other LLM frameworks like LangChain, OpenAI, and more. While leveraging xpander’s powerful agent-building tools, businesses can integrate their AI framework of choice with ease. The xpander-sdk handles function calling and workflow execution in a way that is agnostic to the chosen AI framework, allowing AI Agents to retrieve and run tools seamlessly across systems.
xpander-sdk and xpander’s Agent Graph System enforces structured, dynamic task flows, enabling AI Agents to invoke function calls and execute multi-step workflows without requiring manual management of graph logic, prompt engineering, or runtime orchestration. This flexibility empowers organizations to design intelligent agents that fit into their existing AI ecosystems, all while xpander simplifies the underlying complexity of writing function calling, manage the tool execution and orchestration.