Abstract Opioid drugs like morphine are the gold standard for the treatment of chronic pain, but are limited by adverse side effects, such as tolerance, constipation, and reward/addiction. In our earlier work, we showed that Heat shock protein 90 (Hsp90) has a crucial role in regulating opioid signaling that differs between brain and spinal cord; Hsp90 inhibition in brain blocks opioid pain relief, while inhibition in the spinal cord enhances it. Building on these findings here, we injected the non-selective Hsp90 inhibitor KU-32 directly into the spinal cord of male and female CD-1 mice, showing that morphine anti-nociceptive potency was boosted by 1.9-3.5 fold in the pain models of tail flick, post-surgical paw incision, and HIV peripheral neuropathy. At the same time, morphine tolerance was reduced from 21 fold to 2.9 fold and established tolerance was rescued, while the potency of constipation and reward (as measured by conditioned place preference) was unchanged. These results demonstrate that spinal Hsp90 inhibition can improve the therapeutic index of morphine. However, we also found that systemic non-selective Hsp90 inhibition resulted in a brain-like effect, blocking opioid pain relief. We thus sought a way to circumvent the effects of brain Hsp90 inhibition by investigating the molecular Hsp90 isoforms active in regulating opioid signaling in both regions. Using selective small molecule inhibitors and CRISPR gene editing, we found that 3 Hsp90 isoforms regulated spinal cord opioid signaling (Hsp90α, Hsp90β, and Grp94) while our previous work showed only Hsp90α was active in brain. We thus hypothesized that a systemically delivered selective inhibitor to Hsp90β or Grp94 could selectively inhibit spinal cord Hsp90 activity, resulting in enhanced opioid pain relief and decreased side effects. We tested this hypothesis using intravenous delivery of KUNB106 (Hsp90β) and KUNG65 (Grp94), showing that both drugs enhanced morphine potency in tail flick and paw incision pain while rescuing anti-nociceptive tolerance. We also found that intravenous KUNA115 (Hsp90α) fully blocked morphine anti-nociception. Together, these results suggest that selective inhibition of spinal cord Hsp90 isoforms is a novel, translationally feasible strategy to improve the therapeutic index of opioids.