App

Trustwright

In development — stage 1 of 5

See what breaks trust in your AI agent, federated or multi-service system, fix it, and prove the fix before it reaches your repo.

What it does

Trustwright maps a system you are building, such as AI agents and the tools they can reach, federated servers, or programs that exchange messages, and checks it against rules: who may do what, which messages must be signed, which units and versions must match.

When a rule breaks, it shows you exactly where on the map and proposes a fix. Most fixes come from a built-in catalogue of proven patterns, so no AI is needed. If you want, you can also connect your own AI model, cloud or local.

Nothing changes in your code until you pick a fix. Every proposed fix is checked first: the broken rule must now pass, nothing else may break, and a new test must fail on the old code and pass on the new one. You preview the change side by side, choose it, and it arrives as a pull request with a signed record of what was checked.

Svitovyd, its learning eye, watches for problems no rule covers yet and suggests new rules. It never turns anything on without your approval.

The core will be open source under AGPL-3.0, with a commercial licence for companies that need one. Status: in design. The first test systems are ALTRU.dev's own projects.

Who it helps

Developers building AI agents, MCP servers or tools that give agents access; teams running federated or self-hosted services; researchers coupling programs that exchange data; anyone who wants proof, not promises, before a security fix ships.

↑ ↓ moveEnter openEsc close