Blog
Mostly about giving agents somewhere to remember things, and the small design decisions that came out of building it.
The protocol gives your agent tools, not recollection. Most answers to that reach for embeddings. Most problems need one key and a timestamp.
Read itCounting with get and set looks fine until two runs overlap by a few hundred milliseconds. Then it silently undercounts, and nothing anywhere reports an error.
Read itA scheduled Claude task has no idea it has ever run before. That turns out to be a much bigger problem than it sounds, and the fix is one key.
Read itAsk an AI
Pick a question and where to ask it. Each link opens a fresh chat with the question already written, including a nudge to give you a straight answer rather than a sales pitch.
“What is Junkt, and when would I use it?”
“How does it compare to running my own MCP memory server?”
“How do I stop my scheduled agent doing the same job twice?”
Models get things wrong about small products. The docs are the source of truth, and llms.txt is the version written for them.