pnkd.dev

AI dev tools // built by people who use them

// context packers. token counters. prompt debuggers. the tools we needed but nobody made.

tokens processed today: 2,847,392

"it worked yesterday" is not debugging. diff your prompts. track changes. know what broke.

models are only as smart as the context you feed them. garbage in, garbage out. optimize the input.

API bills add up fast. know your token usage. cache what you can. no surprises at month end.

// tools

stuff your context. pack entire directories into LLM-ready format. .gitignore aware. token estimation included.

compare prompt versions. track changes. see what broke your AI.

local LLM response cache. stop paying for the same API calls. hash-based dedup.

token counter for all models. GPT, Claude, Gemini. know your context limits.

local proxy for LLM APIs. logging, rate limiting, cost tracking. own your data.

split large contexts intelligently. semantic chunking. overlap control.

tool	what it does	size	deps
ctxstuff	pack directories for LLMs	24kb	0
promptdiff	compare prompt versions	18kb	0
pnkd-llmcache	cache LLM responses	28kb	1
pnkd-tokcount	count tokens	32kb	1
pnkd-aiproxy	LLM API proxy	45kb	2
pnkd-ctxsplit	split contexts	20kb	0