Own your agent security — don't trust agents by default.
An agent is only as safe as the layer between it and your machine. A firewall that gates every tool call, a supply-chain gate that vets every skill before it loads, a vault that hands out leases, never keys. Three open-source tools, one layered defense — built here, run in production here.
You don't rent your servers by the request. You don't rent your database by the row. So why rent your intelligence by the token?
Thirteen tools. One machine you own.
Not a pile of repos — a coherent stack, each layer open source and useful on its own. Every model call flows up through a router you own, gets real work done, and passes a security layer that gates every action. Pick one tool, or run the whole thing.
Each tool is MIT-licensed and useful on its own — the full set lives at github.com/askalf. On top of all of it sits askalf — the AI operation that runs Sprayberry Labs on this exact stack, one human approving what matters.
Own one layer first.
You don't adopt a stack on day one. Each tool is a door — start at the one that fixes your sharpest problem, and the rest are there when you need them.
This isn't a concept. The whole stack runs Sprayberry Labs — the software studio with one human on staff — in production today, via askalf, its AI operation: routing every model call, researching, browsing, guarding each action, and shipping the work. The tools are open because they were built to be used, not pitched.
Build on ground you own.
Pick a tool. Read the build notes. Run it on your own box.
Explore the stack →