What is an Agent?

An agent is the main building block of the PrimePilot platform. An agent takes a message or a file as input and acts upon that.

Agent Types

PrimePilot supports several types of agents, each designed for different use cases and capabilities.

Proxy Agents

The most simple agent is a proxy agent to an LLM such as ChatGPT or Anthropic. A proxy agent does nothing additional beyond passing user messages back and forth to the underlying LLM.

Having proxy agents, even without additional capabilities, is helpful because we can share the same context between many different proxy agents. This allows for consistent conversation history and context management across multiple agent instances.

RAG Agents

The main feature of PrimePilot is building, training, and maintaining agents with additional capabilities for long-term specific use cases. For that, you can have RAG (Retrieval-Augmented Generation) agents.

A customer service agent or a domain expert agent are examples of RAG agents. These agents are trained on your knowledge base and can provide intelligent responses based on your specific content.

An unconfigured RAG agent acts as a proxy agent to the underlying LLM. Once you add knowledge sources and configure the agent, it becomes a specialized RAG agent capable of answering questions based on your documentation and data.

MCP Agents

MCP (Model Context Protocol) agents wrap MCP servers. With just a simple configuration, we can link any MCP server to PrimePilot.

Many popular SaaS products such as Asana, Stripe, Xero, and others offer MCP servers and can easily work with PrimePilot agents. This allows you to integrate external services and tools directly into your agent workflows.

Getting Started

To create your first agent, navigate to the Agents section in your dashboard and choose the type of agent that best fits your needs. You can start with a simple proxy agent and gradually add capabilities, or configure a RAG agent with your knowledge base from the beginning.