mcp-programmatic-tool-ca-when-a-task-requires-chaining-multiple-tool-calls--c4671be9

THEN Implement programmatic tool calling ('code mode'): convert MCP tool schemas into typed API functions inside a sandbox; let the model write and execute a script that composes tool calls; return only the final result to the model. This avoids sending intermediate results back to the model.

```typescript
// Auto-generated from MCP tool schema
function logging_getLogs(input: {
  level: "error" | "warn" | "info";
  since: number;
}): Promise<{ entries: LogEntry[] }> {
  return mcp.callTool<{ entries: LogEntry[] }>("logging_getLogs", input);
}
```