Build Agentic Integrations

An agentic integration is where you build your agents. It holds the agents themselves, the entry points that let users and other systems reach them, and the model providers, connections, and functions they use. A project can hold several agentic integrations.

Open an agentic integration to see its overview. The overview diagram shows each entry point and the agent it calls, and has Add Agent and Add MCP buttons. The sidebar lists everything in the agentic integration under Agents, Entry Points, Model Providers, Connections, Types, Functions, and Configurations.

What you can add

Add	What it does	Where to add it	Learn more
Agent	An LLM-driven agent with a role, instructions, a model, tools, and memory.	Overview > Add Agent	Create an agent
Trigger	Lets users and systems reach an agent: chat channels such as Agent Chat, Telegram, or WhatsApp Business, HTTP API endpoints, event sources such as Kafka or Salesforce, and file sources such as FTP / SFTP. Each trigger appears under Entry Points in the sidebar.	Agent canvas > Add Trigger	Expose the agent as a chat endpoint
Tools	Functions, connector actions, other agents, knowledge bases, and MCP servers the agent can call.	Agent canvas > Add Tool	Give tools to an agent
MCP server	Exposes your tools to AI assistants such as Claude Desktop or GitHub Copilot over the Model Context Protocol.	Overview > Add MCP	MCP integration
Knowledge	Grounds agent answers in your own documents through retrieval-augmented generation (RAG).	Agent canvas > Add Tool > Use Knowledge Base	Knowledge Bases
Building blocks	Model providers, embedding providers, vector stores, knowledge bases, chunkers, and data loaders shared by the features above.	Created from the agent, tool, and knowledge base forms	AI building blocks

What's next

• AI Integrations: agents, knowledge bases, MCP, and the building blocks behind them.
• Create an agent: add your first agent and chat with it.