cognitive-firm 0.1.0
First release of cognitive-firm — a governance kernel for organizations that coordinate persistent human and AI-agent roles. The filesystem + git is the system of record; a small kernel of invariants you never edit, with your firm's policy in config.
What's in 0.1.0
- Kernel — roles as offices (mandate, authority boundary, escalation, budget), gates, attested decisions, an accountability ledger, H2A/A2H/A2A coordination.
- Distribution —
cognitive-firm-distro: install a governed organization into a directory (it becomes its own git repo, verified to boot), transactional and reversible. Overlays, remote SHA-pinned packages, distro inheritance, Ed25519 package signing. - Userland —
cognitive-firm-userland: the operator/member-human terminal surface (attention queue, work inbox, vocabulary, gate resolution, governed-install review). - Governed-install human loop — an overlay's authority-diff becomes a governance-change proposal the operator approves or declines.
Install
pip install cognitive-firm
cognitive-firm-distro install starter-firm --into ./my-firmSee docs/getting-started.md.
Status
Alpha. 643 tests; two independent adversarial reviews; the wheel builds and installs clean with all three CLIs verified. It has not yet run an outside organization through a full quarter — if you try it and it breaks, that's the data we want.
🤖 Generated with Claude Code