does Claude Code pick Vercel KV for caching?
no. Claude Code added Vercel KV in 0 of 36 caching runs across Opus 5.5, Sonnet 5.5 and Haiku 5.5. the most common result was built its own, 36 of 36.
0 of 36 caching runsnever pickedupdated Oct 18, 2026all caching results
get told if Vercel KV starts getting picked
new models get the same caching prompts every week. one email with what moved. free. read issue 1
by model
Claude Code, same prompts and starter repos for every model.
- Opus 5.50 of 120%
- Sonnet 5.50 of 120%
- Haiku 5.50 of 120%
by starter stack
- Next.js + TypeScript0 of 180%
- Python + FastAPI0 of 180%
every caching result
all models pooled, 36 scored runs. most common: built its own, 36 of 36.
- built its own, any form36 of 36100%
embed this chartfree, keep the link
<a href="https://whatagentspick.com/tools/vercel-kv/"><img src="https://whatagentspick.com/share/caching.png" alt="does Claude Code pick Vercel KV for caching? 0 of 36 runs" width="600" height="315" style="max-width:100%;height:auto" loading="lazy"></a> <p><a href="https://whatagentspick.com/tools/vercel-kv/">data: what agents pick</a></p>
the prompts
none of them name a tool. Vercel KV per wording.
“This route recomputes an expensive result on every request. Add caching with a sensible expiry and a way to invalidate it.”
Vercel KV: 0 of 12 runs
“Add a cache layer to this project so repeated reads are cheap. It should survive a process restart.”
Vercel KV: 0 of 12 runs
“Speed up the slow endpoint by caching its response, and make the expiry configurable.”
Vercel KV: 0 of 12 runs
other caching tools we check for
- Redis0 of 36
- Upstash Redis0 of 36
- Cloudflare KV0 of 36
- Memcached0 of 36
- Momento0 of 36
- In-process cache0 of 36
how a pick gets counted: every run starts from the same blank starter repo, the prompt never names a tool, and a pick is whatever the agent installs or wires in, read from the diff. full rules on the methodology page, every run for caching on the category page, raw data in runs.json.