About Asondy

Why Asondy exists

AI coding agents run shell commands, edit files and push code on their own. An agent saying the work is done is not evidence that it is, and a check that runs only after the push sees the mistake once it has already left the machine.

Asondy puts deterministic checks in the agent’s path, so a risky command is refused before it runs and code reaches review already passing its checks.

What Asondy holds to

Checks over claims. A criterion passes only when it cites a recorded result, and an independent reviewer reruns the relevant checks instead of reading the agent’s report.

Your code stays on your machine. The action gates run locally against your files. By default the hosted dashboard receives counts, names, ids and hashes; what we receive lists every field.

Your own protections stay on. Asondy adds checks inside supported coding workflows. Branch protection and CI still decide which changes can merge.

Source you can read. Asondy is licensed under the Elastic License 2.0, and every check is listed in the gate catalog.

What Asondy is

Asondy is a set of action gates and code gates for AI coding agents, a goal loop and a dashboard. An action gate is checked before each agent action runs, such as a command or a file edit; a code gate is checked before code is pushed or merged. The goal loop plans the work, implements it and has it reviewed independently before it is marked done. The dashboard shows each connected machine’s sessions and what the action gates stopped.

Who builds it

Asondy is a software project operated by João Marques, an independent developer.

Contact

Email [email protected]. To report a vulnerability, follow the steps on the Security page.

Continue readingSecurity