Integrations

A capable agent is rarely a single model responding to a prompt. In practice, agents are networks of interlocking systems, an LLM reasons and decides, tools act on the world, databases store and retrieve knowledge, and external services provide real-time data. Each layer depends on the others, and the agent is only as powerful as the connections between them.

Railtracks is designed around this reality. Whether you're grounding responses with a vector store, executing code in a sandbox, pulling live data from an API, or delegating work to a sub-agent, each integration plugs into the same node-based model — consistent, composable, and easy to swap out.

The guides in this section cover the integrations Railtracks supports out of the box.