Compliance Studio

Compliance Studio started from a specific annoyance: architecture and coding-standards review kept happening as a separate CI gate — something a human had to remember to run, then wait on, then come back to — instead of something that happened while the code was still open in front of you. I wanted that feedback loop inline, not detached from the moment you're actually making the decision.

So I wrapped a CLI I'd already built for exactly this — seven review agents covering architecture boundaries, coding standards, ATDD, security, UX, observability, and process, running against a live Nx project graph in either a fast --mode pr pass or a full --mode full pass — in a VS Codium extension. The extension doesn't reimplement any of that review logic: it discovers the local LM Host's port through VS Code's extensionDependencies mechanism, spawns the CLI as a child process, and streams results back line by line as each agent finishes, so a finding shows up in the tree view the moment it's ready rather than all at once at the end of a run. Every result also gets written to a local PouchDB audit log, and the tree view stays current by subscribing to that database's live changes feed.

It also registers itself as a chat participant, so you can ask it questions about a specific finding directly in the shared chat sidebar instead of only reading a static report. The result is the same review discipline that keeps this monorepo's own bounded contexts honest, but reusable — packaged as an extension anyone on a team can install, not a one-off script only I know how to run.

flowchart TD
    Editor["Editor (inline)"] --> Agents

    subgraph Agents["Compliance-review agents"]
        Arch["Architecture\nboundaries"]
        Standards["Coding\nstandards"]
        Security["Security"]
        Docs["Documentation"]
    end

    Agents --> Findings["Findings reported\n(same as a PR reviewer)"]