<p align="center" <img src="assets/readme-banner.webp" alt="tama-desktop by Wisent" width="100%" </p <!-- wisent-banner:end --
Tama: Never Get Frustrated by AI Again—Block the Behaviors You Don’t Want
Your AI Agent Made a Mistake? Tama Creates Rules So That It Never Happens Again.
AI Agents are powerful but can be so frustrating. Shouting at the screen because your agent hardcoded credentials somewhere or used inline scripts instead of reusable files is not a productive use of your time. Yet you spend your days correcting your AI, just to see the same mistakes over and over again.
Tama stops this with interactive hook management. We generate and propose new hooks for you, making sure AI Agents stop when they do something bad. We can also work proactively, scanning your codebase for previous violations and fixing issues with it.
Want to deslopify your codebase? Or get rid of the agent handing off to you before it is truly done with a task? Or maybe you just hate it when it tells you a particular word and want to make sure you never see a response formatted like this?
For anyone babysitting AI Agents, Tama is a cradle. Make Sure They Act Within Your Rules.
Problem and intended users
Coding-agent policy is distributed across agent adapters, Git dispatchers, a privileged macOS policy backend, and a versioned hook catalog. Without a single control surface, an operator cannot reliably answer which policy is loaded, whether enforcement is ready, or how to recover when a hook blocks necessary work.
Tama is for:
- Wisent developers running supported coding agents on their own Mac;
- security or platform operators diagnosing local policy state;
- repository maintainers scanning a working tree for violations of the approved pre-write rules.
Tama is not intended for untrusted multi-user Macs or people without authorization to change the local agent-policy installation.
Product boundaries
Tama currently includes:
- a read-only view of the integrity-sealed hook catalog bundled with the app;
- inspection of hook purpose, events, side effects, and justification records;
- inspection of live Claude, Codex, OMP, and compatible supervised sessions;
- explicit per-session hook overrides;
- explicit installation of the local session runtime and macOS policy backend;
- an emergency, confirmed disable/re-enable operation for Tama-managed hooks;
- read-only repository violation scans and an explicitly confirmed agent-assisted cleanup workflow.
- first-use and Settings actions that adopt an existing Tama policy bundle into
the editable application-data store without activating it;
Explicit non-goals: