Skip to content

How It Works

What are Agents?

Our AI-powered Agents analyze your Lime CRM data and return structured, predictable outputs. Unlike general-purpose AI chatbots, Agents are configured with:

  • Specific Instructions - Tell the agent what to analyze and how to think about it
  • Defined Input - Specify which Lime object type and properties the agent should examine
  • Additional Context - Provide supplementary data through Lime Queries
  • File Input - Include files referenced by file properties (such as images or PDFs) as part of the agent's input
  • Structured Output - Define exactly what parameters the agent should return, with specific data types

This structured approach ensures that agents return consistent, validated results that can be reliably used in automations or to support user decisions.

Using Agents in Automations

Agents can be used as steps in Lime Automations workflows.

  1. Add Agent Step - In the automation builder, add an "Agent" action step
  2. Select Agent - Choose which configured agent to run
  3. Select Input Lime Object - The Lime object that will be sent to the agent
  4. Use Outputs - The agent's output parameters become available to subsequent automation steps

Files as Input

When the input object or an additional data query includes a file property, the content of the referenced files is sent to the agent alongside the rest of the data. This lets an agent reason over images, PDFs, and other documents attached to your Lime objects.

Support for specific file formats and limits on file size, page count, and the number of files per request depend on the selected model. When Lime Admin detects that an agent configuration includes file properties, a banner is shown on the agent's configuration page, listing the formats and limits that apply.

Output Types

Agents support several output parameter types, each returning a specific datatype:

Text (str)

Text output.

Example: "This lead shows high purchase intent based on multiple website visits and demo request."

Number (int)

Integers only.

Example: 5

Decimal (float)

Decimal numbers.

Example: 0.87

True/False (bool)

True or false values.

Example: true

Predefined Options (enum)

One value from a predefined list of allowed options.

Example:

  • Allowed options: "volvo", "bmw", "porsche", "mitsubishi"
  • Agent returns: "porsche"

Lime Object (limeobject)

A Lime object (e.g., a person, company, or deal).

Example:

  • Configuration: limetype = "person"
  • Agent returns: a person Lime object (or null)