# kea

> A second brain that checks its facts, for people and their AI agents. Plain Markdown for sourced facts, decisions
> with revisit dates, predictions that get scored (Brier score and calibration), a roadmap, and plans reviewed by a
> council of independent AI models; plus a local SQLite FTS5 index over past Claude Code, Codex and OpenCode
> sessions, ChatGPT and Claude exports, git history and notes. Python 3.15, no dependencies, MIT.

Install: update uv (`uv self update`, or `brew upgrade uv` or `pip install -U uv`), then
`uv tool install --python 3.15 kea-brain` (PyPI), `kea init ~/brain` and `kea skill install`.
Agents: the `kea mcp` server, three skills (kea, kea-council, kea-frameworks), and a Claude Code plugin
(`/plugin marketplace add jpsanders/kea`, then `/plugin install kea@kea`). Every command takes `--json`.

## Docs

- [README](https://github.com/jpsanders/kea/blob/main/README.md): what kea is, install, agents, privacy
- [Manual](https://github.com/jpsanders/kea/blob/main/docs/MANUAL.md): every command; the same text as `kea guide`
- [Skill: kea](https://github.com/jpsanders/kea/blob/main/skills/kea/SKILL.md): how an agent works with a brain
- [Skill: kea-council](https://github.com/jpsanders/kea/blob/main/skills/kea-council/SKILL.md): a council review
- [Skill: kea-frameworks](https://github.com/jpsanders/kea/blob/main/skills/kea-frameworks/SKILL.md): adding a lens
- [AGENTS.md](https://github.com/jpsanders/kea/blob/main/AGENTS.md): working on kea's own code
