Skip to main content

Understanding tools in the toolbox

Updated this week

When you configure a responsibility in CollegeVine’s new agent platform, you will see that the Agent has access to specific tools. If you think of a responsibility as a job, tools are like the picks and shovels it uses to get that job done. In the context of CollegeVine, tools allow your AI agent to take action, such as sending an email, setting a reminder, or looking up a document.

This article is your reference guide to the Toolbox. It explains what each tool does, when it is typically used, and how it fits into a responsibility.


What is a tool?

A tool is a single, well-defined action that an AI agent can use while carrying out a responsibility. Agents decide what to do. Tools help define how to do it once a decision has been made.

For example, an agent with a FAFSA follow-up responsibility may decide that a student needs a reminder. The agent determines whether that reminder should be sent by email, text message, or phone call and then calls on the relevant tool to do so.

Tools further allow CollegeVine to expand what agents can do over time.


Available tools

Below is a description of each tool you may see when configuring a responsibility.

  • browser_agent

    What it does: Allows an agent to interact with external websites or portals through automation.

    When it’s useful:

    • Checking or updating information in third-party web portals

    • Completing steps or forms on behalf of a constituent

    • Scheduling to a calendar on behalf of a constituent

  • call_constituent

    What it does: Schedules an outbound phone call to a constituent from the AI agent.

    When it’s useful:

    • Time-sensitive outreach

    • Complex conversations that benefit from voice interaction

    • Follow-ups that require confirmation or clarification

    Calling windows and time zones are respected to ensure appropriate and compliant outreach.

  • cancel_alarms

    What it does: Cancels one or more previously scheduled alarms by name.

    When it’s useful:

    • Stopping future reminders once a task is completed

    • Cleaning up follow-up actions that are no longer needed

    This tool is typically paired with the set_alarm tool.

  • constituent_asset_metadata

    What it does: Reads information about files and documents associated with a constituent, without accessing the file contents.

    When it’s useful:

    • Checking whether a transcript, essay, or recommendation exists

    • Confirming document status or basic details

    • Determining what information is available before taking next steps

    This tool is read-only and does not modify any data.

  • constituent_knowledge

    What it does: Reads or updates structured information about a constituent, such as known attributes, statuses, or flags.

    When it’s useful:

    • Reading existing data to personalize responses

    • Recording outcomes of interactions

    • Updating known information after verification

    Access is governed by permissions to ensure only appropriate properties are changed.

  • direct_llm_call

    What it does: Runs a targeted AI prompt to generate an answer or piece of content.

    When it’s useful:

    • Generating summaries or explanations

    • Answering complex questions using structured inputs

    • Producing content that does not trigger follow-up actions

    This tool returns information only and does not contact constituents or trigger workflows.

  • email_constituent

    What it does: Sends an email to a constituent.

    When it’s useful:

    • Routine outreach and updates

    • Responding to incoming emails

    • Sharing instructions, links, or next steps

    Emails can be fully written by the responsibility or generated with AI assistance, depending on configuration.

  • escalate

    What it does: Hands a case off to human staff for review or follow-up.

    When it’s useful:

    • Sensitive or complex situations

    • Requests that require human judgment

    • Issues outside the agent’s defined scope

    The constituent is notified that their request has been escalated.

  • get_processed_document

    What it does: Retrieves results from a document that has already been processed by the system.

    When it’s useful:

    • Checking the status of document analysis

    • Accessing extracted information from uploaded files

    This tool does not start processing. It only retrieves results.

  • get_transcript_summary

    What it does: Generates a structured summary of a student transcript based on prior document processing.

    When it’s useful:

    • Quickly reviewing GPA and academic details

    • Supporting advising or verification workflows

    This tool returns data only and does not trigger outreach.

  • knowledge_base

    What it does: Searches your institution’s Knowledge Base for answers.

    When it’s useful:

    • Answering policy or program questions

    • Providing accurate, school-specific information

    • Grounding responses in approved content

    Responses are drawn directly from your maintained Knowledge Base.

  • planning

    What it does: Creates a structured plan for how an agent should proceed, based on context and objectives.

    When it’s useful:

    • Multi-step workflows

    • Situations with multiple possible next actions

    • Complex decision-making scenarios

    This tool supports strategy and sequencing rather than execution.

  • process_document

    What it does: Starts a document-processing workflow on a stored file.

    When it’s useful:

    • Extracting data from transcripts, forms, or essays

    • Preparing documents for later review or summarization

    Processing happens asynchronously and can be checked later with get_processed_document.

  • retrieve_sensitive_properties

    What it does: Retrieves sensitive constituent information after confirming identity.

    When it’s useful:

    • Accessing protected or authenticated data

    • Ensuring privacy and compliance before sharing information

    If authentication is required, the constituent is prompted before data is returned.

  • set_alarm

    What it does: Schedules a time-based reminder for future action.

    When it’s useful:

    • Follow-ups after a set period

    • Deadline reminders

    • Time-based nudges

    Alarms can be managed or cancelled later using cancel_alarms.

  • text_constituent

    What it does: Sends a text message (SMS) to a constituent.

    When it’s useful:

    • Short, time-sensitive messages

    • High-engagement reminders

    • Quick confirmations or nudges

    Messages can be written directly or generated with AI assistance.


How to use this guide

When building or reviewing a responsibility, use this article to:

  • Understand what each tool is capable of

  • Decide which tools a responsibility truly needs

  • Avoid overloading responsibilities with unnecessary capabilities

Granting the right tools helps your AI agents stay focused, compliant, and effective.

Next steps

If you’re unsure which tools to use for a specific responsibility, or if you have questions about best practices, reach out to your CollegeVine partnerships team for guidance.

Did this answer your question?