What Is an Agent?

Instead of responding randomly, agents operate with:

• A specific responsibility

• Clear instructions

• Access to approved knowledge

• Controlled behavior

Agents are designed to produce reliable outcomes, not just conversations.

What Agents Are Used For

Agents are typically used to:

• Answer customer or internal questions

• Make decisions based on context

• Perform structured tasks inside workflows

• Act as intelligent interfaces between users and systems

Each agent is optimized for one primary job.

Agent Components

Every agent in Sprig is made up of four core parts.

1. Role

The role defines what the agent exists to do.

Examples:

• “Customer support assistant”

• “Lead qualification agent”

• “Internal operations helper”

A clear role improves accuracy and predictability.

2. Instructions

Instructions define how the agent should behave.

Good instructions are:

• Specific

• Bounded

• Aligned to one task

Example:

You are a customer support agent.
Only answer using official documentation.
If unsure, ask for clarification or escalate

3. Knowledge Access

Agents don’t guess — they reference approved knowledge.

You can connect agents to:

• Product documentation

• Help articles

• Internal notes

• External sources

Example configuration:

{
  "knowledge_sources": ["help_center", "product_docs"],
  "retrieval_depth": "high"
}

Common Mistakes

• Overloading one agent with multiple roles

• Giving vague or open-ended instructions

• Connecting too many data sources

• Treating agents like general-purpose chatbots

Agents work best when scoped tightly.