Skip to main content
Tools transform your knowledge agents from passive information providers into active problem solvers. By integrating specialized capabilities, your agents can perform tasks, retrieve real-time data, make calculations, and connect to enterprise systems—all while maintaining the conversational experience users expect.

What Are Tools?

In Prisme.ai, tools are specialized functions that extend an agent’s capabilities beyond retrieving and presenting information from your knowledge base. They allow agents to:

Perform Actions

Execute tasks based on user requests

Access Live Data

Retrieve real-time information from external systems

Process Information

Perform calculations, transformations, and analysis

Integrate with Systems

Connect with enterprise applications and services

Tool Types

Prisme.ai supports several types of tools, each with specific purposes and capabilities:
  • Native Tools
  • Custom Code Tools
  • API Connector Tools
  • Webhook Tools
  • Database Tools
Built-in tools provided by Prisme.ai, ready to use with minimal configuration.

Common Tools

Prisme.ai includes a growing library of native tools that you can add to your agents with minimal configuration:
Perform mathematical calculations including arithmetic, algebra, statistics, and unit conversions.Capabilities: Complex math operations with memory and variablesUse Cases: Financial calculations, engineering formulas, data analysis
Create custom images from text descriptions using AI generation capabilities.Capabilities: Generate images from text prompts with style controlsUse Cases: Visual content creation, concept visualization, design assistance
Generate alternative phrasings of user questions to improve search results.Capabilities: Create semantic variations of input questionsUse Cases: Search enhancement, query expansion, intent clarificationExample Implementation:
- MyAgent.chat-completion:
    messages:
      - role: assistant
        content: Given the following user question, generate 3 alternative phrasings that express the same intent. Return the result as a JSON object with a "reformulations" key.
      - role: user
        content: 'Question: "{{user_question}}"'
    output: reformulations
- set:
    name: output
    value:
      value: "{{reformulations}}"
      description: alternative phrasings of user question

Creating Custom Tools

You can create custom tools to address specific needs in your organization:
1

Define Tool Requirements

Determine what the tool should do and how it should interact with users and systems.Consider:
  • What problem does this tool solve?
  • What inputs will it need?
  • What outputs should it provide?
  • Which systems will it interact with?
  • What security considerations are important?
2

Select Tool Type

Choose the most appropriate tool type based on your requirements.Tool selection guidance:
  • Use native tools for common functions
  • Choose custom code for specialized logic
  • Select API connectors for external service integration
  • Use webhooks for complex process orchestration
  • Implement database tools for structured data queries
3

Configure or Develop the Tool

Set up the tool’s parameters or develop its code.For custom development:
  • Create clear input/output schemas
  • Implement robust error handling
  • Ensure appropriate security measures
  • Test with various input scenarios
  • Document the tool’s functionality
4

Test the Tool

Verify that the tool works as expected in different scenarios.Testing should include:
  • Basic functionality verification
  • Edge case handling
  • Performance assessment
  • Security testing
  • Integration testing with agent
5

Add to Your Agent

Integrate the tool with your knowledge agents.During integration:
  • Provide clear tool descriptions
  • Configure when the tool should be used
  • Set appropriate permissions
  • Define error handling behavior
  • Test the agent with the new tool

Tool Usage Strategies

Effective tool integration requires thoughtful configuration of when and how tools are used:
Configure how the agent decides which tool to use in different situations.
Determine when the agent should ask for permission before using tools.
Configure how tool results are presented to users.
Define behavior when tools fail or produce unexpected results.

Integration Examples

  • Customer Support
  • Financial Advisor
  • HR Assistant
  • Research Assistant
Agent Purpose: Answer product questions and resolve customer issuesIntegrated Tools:
  • Ticket Creator: Create support tickets for complex issues
  • Order Lookup: Check order status and details
  • Product Catalog: Access current product information
  • Document Analyzer: Process customer documents and screenshots
  • Email Composer: Send confirmation and follow-up emails
Benefits:
  • Streamlined issue resolution
  • Reduced handoffs between systems
  • Consistent information across channels

Security and Governance

Ensuring secure and responsible tool usage is critical:

Authentication & Authorization

Configure secure access to tools and their underlying systems.Key Controls:
    Tool-specific permissionsAPI key managementCredential rotationUser-based access control

Data Privacy

Ensure tools handle sensitive information appropriately.Key Controls:
    Data minimizationInformation redactionTemporary storage policiesPrivacy-preserving integration

Logging & Auditing

Maintain records of tool usage for compliance and security.Key Controls:
    Comprehensive activity logsTool usage analyticsAudit trailsAnomaly detection

Rate Limiting & Quotas

Prevent abuse and manage resource consumption.Key Controls:
    Usage quotasRate limitingCost controlsResource allocation

Best Practices

  • Start with native tools when possible for easier maintenance
  • Choose tools that solve specific user problems
  • Consider the full lifecycle of the tool, including updates and maintenance
  • Select the simplest tool type that meets your requirements
  • Prioritize tools that integrate with existing systems
  • Provide clear, specific descriptions of what each tool does
  • Set appropriate input validation to prevent errors
  • Configure sensible defaults for optional parameters
  • Include example inputs to guide users
  • Test tool configuration with various scenarios
  • Make tool usage transparent to users
  • Provide clear feedback when tools are being used
  • Format tool results for readability and understanding
  • Balance automation with user control
  • Implement graceful error handling with helpful messages
  • Regularly review tool usage and performance
  • Monitor for errors and unexpected behavior
  • Update tool configurations as underlying systems change
  • Document tool purpose, configuration, and integration
  • Establish a testing protocol for tool updates

Next Steps

I