Neural-recording ICs have been a key tool to unravel the mystery of the human brain and find treatments for various neurological diseases. Since neural signals inherently have a small amplitude and suffer from environmental interferences, conventional neural recording circuits have been mainly designed for low noise, high CMRR, and low power, using the structure with a high-gain amplifier and a low-resolution ADC [1] (Fig. 1). With the advent of closed-loop neurotherapeutics, stimulation artifacts have been a notorious obstacle in neural recording. To tackle this issue, a direct-conversion structure has been widely used due to its wide dynamic range [3] -[7]. However, the structure could not meet the bandwidth (BW) requirement of 5kHz and the input-referred noise (IRN) requirement of 7 \mu V{rms} simultaneously. In this paper, we present a closed-loop neural-recording IC using an adaptive automatic gain controller (AGC) and continuous-time dynamic-zoom \Delta \Sigma ADC (CT-Zoom-ADC). By combining the AGC and CT-Zoom-ADC, the IRN performance is improved to 6.1 \mu V{rms} at 5kHz BW, and the saturation issue of the conventional amplifier-based recording structure is alleviated. Also, the recording IC can rapidly recover the signal from transient artifacts thanks to the digital auto-ranging block (DAR).