Spec Kitty

Spec Kitty is a decision layer for teams that build software with autonomous coding agents. Its premise is that agents make thousands of small product and engineering decisions at machine speed, and that standards and past rulings usually sit in people's heads or in wikis the agents never read. Spec Kitty gives agents that knowledge and checks their work against it. The product has three parts. Brain Kitty is a personal knowledge graph of the standards, runbooks, rules and decisions a person owns; answers cite their source, and people approve every change. Team Kitty is a hosted team product, currently in early access, that combines each person's graph so the team decides together what applies, then serves it to coding agents over the Model Context Protocol. Build Kitty is the Spec Kitty command line tool, where agents write a spec, a plan and work packages that the user can edit before any code is written. Each work package runs in its own git worktree so several agents do not collide, and a reviewer other than the author checks each change against the spec. The CLI works inside Claude Code, Codex, Cursor, Copilot and more than twenty other coding agents, using the customer's own models and keys. It is free and open source under the MIT license. Team Kitty is described as the paid team product, but no plan prices are published, and Brain Kitty is offered through a contact form. A demo can be booked from the site.