Guides
Tool calling
Define function tools and return tool results through the common message format.
Tools use the widely supported OpenAI function-tool shape. Native adapters translate this shape when their provider uses a different wire format.
import { completion, type Tool } from "any-llm-ts";
const tools: Tool[] = [
{
type: "function",
function: {
name: "get_weather",
description: "Get the current weather for a city",
parameters: {
type: "object",
properties: {
city: { type: "string" },
},
required: ["city"],
},
},
},
];
const response = await completion({
provider: "anthropic",
model: "claude-sonnet-4-5",
messages: [{ role: "user", content: "What is the weather in Kolkata?" }],
tools,
});
const toolCalls = response.choices[0]?.message.toolCalls ?? [];Execute and return a result
Tool arguments are normalized as a JSON string:
const call = toolCalls[0];
if (call?.function.name === "get_weather") {
const args = JSON.parse(call.function.arguments) as { city: string };
const result = await getWeather(args.city);
const followUp = await completion({
provider: "anthropic",
model: "claude-sonnet-4-5",
messages: [
{ role: "user", content: "What is the weather in Kolkata?" },
response.choices[0]!.message,
{
role: "tool",
toolCallId: call.id,
content: JSON.stringify(result),
},
],
tools,
});
}The library does not execute tools or run an agent loop. Your application remains responsible for validating arguments, authorizing operations, executing functions, limiting iterations, and returning results.
Force or disable tool use
toolChoice: "auto"
toolChoice: "required"
toolChoice: "none"
toolChoice: { type: "function", function: { name: "get_weather" } }Provider support varies. Consult the provider's documentation for model-specific limitations.