Tool Use & Function Calling

Tool use, also known as function calling, lets an AI node invoke external tools or APIs during a conversation. Interlocute supports tool use natively — you configure which tools a node can access, and the runtime handles invocation, result injection, and logging automatically.

Interlocute includes pre-configured tools for web search and data processing. You can also define custom tool schemas that point to your own API endpoints. Tools are registered per node, so different nodes can have different capabilities.

You define a tool schema (name, description, parameters) and associate it with an API endpoint. When the LLM decides to call the tool, Interlocute invokes your endpoint with the specified parameters and returns the result to the LLM for incorporation into the response.

Yes. Every tool invocation is logged with the tool name, parameters, result, and latency. You can set execution policies to restrict which tools a node can use, enforce rate limits, and require specific API key scopes for tool access.

Yes. Tool access is configured per node and can be scoped by API key. You can allow all tools, restrict to a specific set, or disable tool use entirely. Changes take effect immediately without redeployment.

Tool calls are executed inline during response generation. If the LLM decides to call a tool mid-response, the tool is invoked, and the result is incorporated into the streamed output. The client sees a seamless response without needing to handle tool orchestration.

Tool invocations are metered as computation tokens. The cost of the tool call itself (e.g., a web search) is included in your node's usage. External API costs for custom tools are your responsibility, but Interlocute tracks and logs every invocation for visibility.