Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.elementum.io/llms.txt

Use this file to discover all available pages before exploring further.

AI agents in Elementum are conversational assistants tied to your apps. They interpret what people ask for, use tools to work with records and automations, and reach users on channels such as Microsoft Teams or phone calls. First-line support is a typical starting point; the same patterns apply to onboarding, sales, or any process where a guided conversation speeds up routine work. For creating an agent, adding it to an app, and configuring models and instructions in the UI, see Building Agents.

​
Tools

Agents use tools to take action during conversations — creating or updating records, running searches, triggering automations, delegating to sub-agents, or calling external services through MCP. Each tool has a name, description, permissions mode (Run as publisher, Run as current user, or Run as service account), and type-specific configuration. Available tool types:
  • Create Record — Create new records in a specified App, Element, or Task
  • Record Search — Search for existing records based on a query (returns up to 100 results)
  • Update Record — Update fields on an existing record identified by record ID
  • AI Search Records — Semantic search using AI to retrieve records based on natural language
  • Run Automation — Execute an automation with an On-Demand trigger, passing input and receiving output values
  • Run Agent — Delegate a task to a sub-agent in a new conversation
  • MCP — Connect to an external MCP server and invoke its tools
  • Skills — Dynamically discover and run reusable Skills at runtime (no manual tool setup required)

​
Configure Agent Tools

Add and configure tools so agents can perform actions like creating or searching records, running automations, delegating to sub-agents, or connecting to external MCP services.
  1. In the Agent profile page, click the Configure button.
  2. Click + Tool in the Tools section.
  3. Select the tool type.
  4. Complete the configuration. Follow the specific instructions below for the tool type you’re creating.
  5. Save the tool.
Every tool type includes a Tool Start Message field, which controls the live status message users see in chat while the tool is running (e.g., “Searching knowledge base…”). Setting a clear, tool-specific message helps users understand what the agent is doing. Multiple back-to-back tool calls appear as separate messages, and this behavior is supported in both the agent builder preview and deployed app agents.
Agent tool permissions override the agent permissions unless set to agent default.

​
Search Records

Enables the Agent to search for existing records based on a query. Configuration uses a three-step wizard: Basic Info, Filter Conditions, and Output Fields. Step 1: Basic Info
  • Tool Name: Enter a name with no spaces or special characters.
  • Tool Description: Describe when and how the Agent should use this tool. This helps the Agent understand when to invoke it.
  • Tool Start Message: Customize what the user sees or hears when the tool runs.
  • Permissions: Select the execution permissions for the tool. Defaults to agent default.
  • Select Object: Choose the record object to search.
  • Max Results: Set the maximum number of results returned per search (default is 10). Keeping this low prevents overwhelming the Agent.
  • Sort Configuration: Define the default sort order for search results returned by this tool. Click Add Sort and select a field to sort by.
Step 2: Filter Conditions
  • Filter Variables: Define variables the Agent fills in at runtime. Click + Add Variable and provide a variable name, a description with valid values, and whether it is required. Use the link button on a filter row to set a variable as the filter value.
  • Filter Conditions: Add static filter conditions to narrow results. Build rules using a field, operator, and value. Click + Condition to add a row or + Condition Group to add grouped conditions.
Step 3: Output Fields
  • Output Fields: Toggle on the fields from the search results that should be returned to the Agent.
The search returns results similar to a list view search. If more than 100 records are found, the Agent receives a failure message indicating the need for a more refined query.
Provides advanced search capabilities using AI, allowing the Agent to understand and retrieve related data based on natural language. Configuration uses a five-step wizard: Basic Info, Select Source, Configure Fields, Output Fields, and Review and Save. Step 1: Basic Info
  • Tool Name: Enter a name with no spaces or special characters.
  • Tool Description: Describe when and how the Agent should use this tool. This helps the Agent understand when to invoke it.
  • Tool Start Message: Customize what the user sees or hears when the tool runs.
  • Permissions: Select the execution permissions for the tool. Defaults to agent default.
Step 2: Select Source
  • Select Object: Choose the record object to search.
  • Search Service: Select the AI search service to use for this tool.
  • Max Results: Set the maximum number of results returned per search (default is 10). Keeping this low prevents overwhelming the Agent.
Step 3: Configure Fields
  • Search Field Description: Describe how the Agent should use the search functionality.
  • Filter Variables: Define variables the Agent can fill in at runtime. Click + Add Variable to create a variable, then use the link button on a filter row to reference it as the filter value.
  • Configure Search Filters: Add filters to narrow down search results. Build filter conditions using a field, operator, and value. Click + Condition to add a row or + Condition Group to add grouped conditions.
Step 4: Output Fields
  • Output Fields: Toggle on the fields the Agent can return. Available fields are specific to the record object chosen in Step 2. Only enable fields relevant to the Agent’s task.
Step 5: Review and Save Review your configuration across all steps before saving. Once saved, the tool will be available for the Agent to use.

​
Create Record

Allows the Agent to create new records in the system.
  • Tool Name: Enter a name with no spaces or special characters.
  • Tool Description: Describe when and how the Agent should use this tool. This helps the Agent understand when to invoke it.
  • Tool Start Message: Customize what the user sees or hears when the tool runs.
  • Permissions: Select the execution permissions for the tool. Defaults to agent default.
  • Record Object: Choose the record object the Agent will create.
  • Fields to Create: Click + Field to add fields the Agent will populate when creating a record. A record object must be selected before fields can be configured. Each field can be given a name, a description to guide the Agent on what value to provide, and toggled as required.

​
Update Record

Allows the Agent to update existing records.
  • Tool Name: Enter a name with no spaces or special characters.
  • Tool Description: Describe when and how the Agent should use this tool. This helps the Agent understand when to invoke it.
  • Tool Start Message: Customize what the user sees or hears when the tool runs.
  • Permissions: Select the execution permissions for the tool. Defaults to agent default.
  • Record Object: Choose the record object the Agent will update.
  • Identifier Description: Help the Agent understand the unique identifier of the record to update (e.g., record ID or name).
  • Fields to Update: Click + Field to add fields the Agent will update on the record. Each field can be given a name, a description to guide the Agent on what value to provide, and toggled as required.
The Agent is notified whether the update was successful or not, along with reasons for failure, allowing it to attempt corrections.

​
Run Automation

Allows the Agent to execute existing automations that have an On-Demand trigger, enabling complex workflows to be triggered conversationally.
  • Tool Name: Enter a name with no spaces or special characters.
  • Tool Description: Describe when and how the Agent should use this tool. This helps the Agent understand when to invoke it.
  • Tool Start Message: Customize what the user sees or hears when the tool runs.
  • Permissions: Select the execution permissions for the tool. Defaults to agent default.
  • Automation: Choose an existing on-demand automation to run.
  • Input Values: Click + Input to define the values the Agent will pass to the automation. Each input can be given a name and a description to guide the Agent on what value to provide.
  • Output Values: Click + Output to define the values the Agent will receive back from the automation. Each output can be given a name and a description so the Agent understands how to use the returned data.

​
Run Agent

Allows the Agent to delegate tasks to another agent (sub-agent) for specialized processing. Each execution creates a new conversation between the main agent and the sub-agent.
  • Tool Name: Enter a name with no spaces or special characters.
  • Tool Description: Describe when and how the Agent should use this tool. This helps the Agent understand when to invoke it.
  • Tool Start Message: Customize what the user sees or hears when the tool runs.
  • Permissions: Select the execution permissions for the tool. Defaults to agent default.
  • App: Choose the App containing the target agent.
  • Target Agent: Choose the sub-agent to delegate tasks to.
  • Worker Task Prompt: Instructions for how this agent should delegate tasks to the worker agent.
Each tool execution starts a new conversation with the sub-agent. The main agent can provide context and instructions, and results from the sub-agent are returned for continued processing.

​
MCP Tool

The MCP (Model Context Protocol) Tool allows the Agent to connect to external MCP servers and use custom tools, extending the Agent’s capabilities beyond the platform’s built-in tools. Configuration follows a three-step wizard: Connect to MCP Server, Select Tool, and Configure Tool. Step 1: Connect to MCP Server
  • MCP Server URL: Enter the base URL of the MCP server (e.g., https://api.example.com/mcp).
  • Authentication: Select the authentication method — Bearer Token, Basic Auth, API Key, or None.
  • Click Connect to discover available tools on the server.
Step 2: Select Tool
  • Browse and select from the tools available on the MCP server.
Step 3: Configure Tool
  • Tool Name: Enter a name with no spaces or special characters.
  • Tool Description: Describe when and how the Agent should use this tool. This helps the Agent understand when to invoke it.
  • Tool Start Message: Customize what the user sees or hears when the tool runs.
  • Permissions: Select the execution permissions for the tool. Defaults to agent default.
The Agent communicates with the external MCP server to execute the tool. Authentication is handled automatically based on the configured method. Results are returned to the Agent for use in the conversation.

​
Ask User Question Tool

Enable the Ask User Question tool in agent settings to let agents surface structured forms mid-conversation to collect user input. Use agent instructions to control when the tool is called and what questions, options, and required fields are generated. When invoked, a UI form is rendered for the user to complete and return to the agent.
Forms are generated by the LLM at runtime, so some variation in wording, options, and field layout is expected between invocations.
Enable the tool:
  1. In the System Tools section, click the Edit icon next to Ask User Question.
  2. Toggle on Enable.
  3. Click Save.
  4. Update the agent instructions (prompt) to describe the form, including:
    • The questions to ask.
    • The desired answer format for each question (for example, single-select, multi-select, or free-text).
    • Whether each answer is required.
  5. Click Save.
  6. Use the Chat Preview to confirm the experience works as desired.
Example instructions: 
Always use the askUserQuestion tool when gathering information from the user.

Make sure you ask for:
1. The service. Required. Zoom, Cursor, or Figma, no custom input.
2. The platform. Required. Web, iOS, Android, Windows, Mac. Multiple allowed. No free entry.
3. The business use case. Text input. Optional. DO NOT put the text (optional) in the question text.
Avoid instructing the agent to include words like “optional” in the rendered question text—set the required/optional behavior through the instructions instead.

​
Dynamic Dropdowns

Pair the Ask User Question tool with Dynamic dropdowns so the agent can populate dropdown options from live platform data instead of static lists defined in instructions. This keeps choices current as records are added, updated, or removed and lets users search within long option lists. To add a dynamic data source:
  1. When configuring an agent on an object, open the Ask User Question system tool.
  2. In the Dynamic dropdowns section of the popup, click + Add Source.
  3. Choose the target object from the Records to Show dropdown.
  4. Choose the field that is rendered to users in the dropdown.
  5. Choose whether the prompt accepts a single-select or multi-select response.
  6. To add another source, click + Add Source in the top-right corner of the Dynamic dropdowns tile.
  7. Click Save.
User experience: When the agent surfaces a question that uses a dynamic dropdown, users see a searchable picker populated from the configured object. The selected value (or values) is returned to the agent and can be used in the next step of the conversation or written to a record.
Use dynamic dropdowns when the answer must match an existing record (for example, “Which project?” or “Which vendor?”). Use the standard instruction-driven options when the choices are fixed and unrelated to platform data.

​
Copy or Duplicate a Tool

Reuse existing tool configurations instead of rebuilding them from scratch. Copy a tool to another agent in the same app:
  1. Open the agent that has the tool you want to copy.
  2. Hover over the tool and click the More icon.
  3. Click Copy Tool to Agent.
  4. Choose the target agent from the dropdown.
  5. Click Copy Tool.
  6. Navigate to the target agent to manage the copied tool.
Duplicate a tool within the same agent: Hover over the tool, click the More icon, and select Duplicate Tool. This creates a copy on the same agent with an auto-generated name to avoid conflicts — giving you a pre-built starting point to modify rather than configuring from scratch.

​
Deploying Agents

Once configured, agents can run in several contexts:

​
External Agents via App Intelligence

In addition to creating and managing internal agents, Elementum supports integration with external agents configured at the App level through Intelligence settings. You can connect specialized AI capabilities from external providers such as Snowflake Cortex and AWS Bedrock while keeping data access governed by your provider configuration.

​
External Agent Configuration

External agents are configured at the App level through Intelligence settings, not centrally. This approach ensures:
  • App-Specific Context: Agents access only the data relevant to their App
  • Scoped Permissions: Security controlled through AI Provider credentials
  • Independent Configuration: Each App can configure agents according to its needs
  • Flexible Deployment: Different Apps can select different providers and agents
Key Concepts:
  • Native agents: Built and hosted in Elementum
  • Managed agents: External agents (for example Cortex or Bedrock) integrated through App Intelligence
  • App Intelligence: Configuration area where external agents are discovered and connected
  • AI Provider: Provider configured with appropriate credentials for agent discovery and invocation

​
Supported external agent runtimes

Elementum supports two external agent runtimes, each documented end-to-end on its own page:

Snowflake Cortex Agents

Run Cortex Agents on your Snowflake data warehouse and invoke them from App automations

AWS Bedrock Agents

Connect a Bedrock Agent built in your AWS account via the Agent Alias ARN
Both runtimes use the same App-level configuration pattern: select the AI Provider you’ve already connected, register the external agent, then invoke it from automations using Run Agent Task. For the full setup, security model, and integration architecture for each runtime, follow the links above.

​
Agent-to-Agent Protocol (A2A)

Elementum supports the Agent-to-Agent (A2A) protocol, an open standard that enables external systems to communicate with Elementum agents programmatically. This allows other AI agents, automation platforms, or custom applications to interact with your Elementum agents as remote collaborators. For the full A2A reference — architecture, authentication, Agent Cards, JSON-RPC examples, streaming, task states, and multi-turn conversations — see Agent Architecture & A2A Protocol.

​
Next Steps