Build a Reliable Tool-Calling Agent Loop on OpenRouter

A tool-calling agent loop repeatedly sends the conversation and available tools to a model, executes any tool calls the model requests, appends the results, and asks the model what to do next. The loop ends when the model returns no tool calls or when an application-defined stop condition fires. The model decides which tool to request, but your application parses the arguments, runs the function, and decides when the loop must stop. This guide shows you how to build that loop with the OpenRouter TypeScript SDK. The example uses local weather data, so you can run it without setting up another service. The model requests tools. Your application executes them and returns one result per tool call ID. Send the tool definitions on every request, including follow-up requests after tool results. Stop when the model returns no tool calls, when the same call repeats past a limit, or when an iteration cap is reached. The models parameter gives you ordered model fallback, and Auto Exacto reorders providers on tool-calling requests by default. Call the model with the full history and tool definitions. If the response has no tool calls, return the assistant’s text and stop. If this is the last…