Creating AI tools
Tools are callable functions that extend what AI agents can do. Each tool combines an edge function (the code that runs) with a JSON schema (describing its inputs), allowing agents to call them during conversations.
What is a tool?
Section titled “What is a tool?”A tool is an edge function deployed to the Quant CDN edge network, registered with a JSON schema so agents know how to call it and what parameters to provide. When an agent determines it needs to perform an action — such as fetching data, processing content, or interacting with an external service — it invokes the appropriate tool with the required parameters.
Tool components
Section titled “Tool components”Each tool is defined by a set of fields that control its identity, behaviour, and execution.
| Field | Required | Description |
|---|---|---|
toolName | Yes | Unique identifier (alphanumeric, hyphens, underscores) |
description | Yes | What the tool does (shown to agents) |
edgeFunctionUrl | Yes | HTTPS URL of the deployed edge function |
inputSchema | Yes | JSON Schema object defining accepted parameters |
category | No | Organisation category (e.g. data, content, ai-generated) |
responseMode | No | direct (raw output) or llm (AI-formatted response) |
isAsync | No | true for long-running operations (agent polls for result) |
timeout | No | Maximum execution time in milliseconds |
Response modes
Section titled “Response modes”The responseMode field controls how the tool’s output is returned to the user.
direct— Returns the tool’s raw output directly to the user. Use this when the tool already produces human-readable content or structured data that the agent should pass through unchanged.llm— Passes the tool’s output through the AI model for natural language formatting before returning to the user. Use this when the raw output needs summarisation, explanation, or conversational formatting.
Async tools
Section titled “Async tools”For long-running operations, set isAsync: true. The agent will start the tool execution and then poll for the result rather than waiting synchronously. This is useful for tools that perform complex computations, call slow external APIs, or process large datasets.
Categories
Section titled “Categories”Organise your tools into categories for easier management. Assign a category when creating a tool — for example, data for data-fetching tools or content for content manipulation tools. Filter by category on the AI Tools page to quickly find the tools you need.
Viewing tools
Section titled “Viewing tools”Navigate to AI Tools in the dashboard sidebar to see all registered tools. The tools list shows each tool’s name, description, category, and deployment status. Use the category filter to narrow the list, or search by name to find a specific tool.
Next steps
Section titled “Next steps”- AI-assisted tool creator — Generate tools from natural language descriptions using AI
- Deploy & test — Deploy your tools and test them before assigning to agents