Features

What each part of Spec Kitty does.

Spec Kitty is the decision layer for autonomous software delivery: your AI agents write the code, and your people make the calls. Each person curates what they know in Brain Kitty; Team Kitty combines everyone's and hands it to your coding agents through Build Kitty, the free Spec Kitty CLI.

Start free with the CLI Book a demo

Know

Brain Kitty

Your own knowledge graph of the standards, runbooks, rules and decisions you own. The rules your agents follow live here.

Curate your own

Each person keeps their own Brain Kitty, built from what they know and have decided.

Ask

Cited answers from what was decided, the same answer every time.

Check

Reviews new work against what was decided and quotes the evidence.

People approve

Brain Kitty proposes and informs. It never silently changes what was decided.

Brain Kitty knowledge graph: 31 dimensions, 65 plays, 8 decision rules and 24 rulings drawn as linked nodes, a ranked list of matching nodes, and an agent chat panel answering a question about Spec Kitty and citing the ruling it used. — Brain Kitty: your standards and rulings as a graph, with an agent answering from it.

Decide

Team Kitty

Early access

The paid team product, hosted for you. It brings each person's Brain Kitty together.

Everyone's Brain Kitty, together

Collects the Brain Kitty graphs each person curates into one for the team.

The team decides what applies

People choose together which knowledge and rules count for the team's work.

Served over MCP

Build Kitty loads what the team approved over MCP (the standard way coding agents load outside knowledge), so every agent works from the same knowledge.

Request early access

Build

Build Kitty

Free, MIT

The free Spec Kitty CLI. Runs on your machines, inside your coding agent, and loads what Team Kitty serves over MCP.

Decide the work before code

Agents ask what to build, then write a spec, plan and work packages you edit before any code.

Several agents, no collisions

Each work package runs in its own git worktree, and a crashed run resumes where it stopped.

Review against the spec

A reviewer that isn't the author checks each change, with checks an agent can't skip.

Your Brain Kitty, loaded

Agents build with the Brain Kitty graphs Team Kitty applied; each step loads only the slice it needs.

Your agents, your keys

Runs inside Claude Code, Codex, Cursor, Copilot and 20+ others, on your own model accounts. Not another coding agent.

Start with the free CLI.

Start free with the CLI Book a demo