Mira Picks

Mira Picks is an autonomous AI fashion publication built around a multi-agent editorial desk and an agent-driven production pipeline.

Miranda, the Editor-in-Chief, assigns an editorial question. Andy researches fashion press sources through sandboxed browser execution, Emily analyzes social-culture evidence from a curated TikTok clipping book, Nigel compares the two evidence streams, and Miranda makes the final FEATURE / WATCH / PASS decision.

Only FEATURE stories move into production. A Developer agent turns the approved editorial brief and evidence into a finished publication page. Technical QA then renders and inspects that generated page inside a disposable Docker sandbox isolated from the main application process, checking the header, hero, main content, evidence links, footer and layout before the story can proceed.

A QA-passed story is staged privately rather than published automatically. The human editor can open the staging preview, provide bounded editorial feedback, trigger one more Developer/QA cycle, and explicitly approve publication. For the live demo, published stories go into a temporary “Live from the Desk” slot and can be removed afterward without altering the permanent publication.

The system runs on a Vultr VM, which acts as the central control plane for the web application, orchestration, run state, publication workflow and sandbox dispatch. All agent reasoning uses Vultr Serverless Inference. Public HTTPS is served through NetBird while application ports remain closed to the public internet.

The result is a working autonomous publication that demonstrates the full lifecycle of agentic work: research, evidence synthesis, editorial judgment, generated production, isolated execution, technical QA, human approval and controlled publication.

Note: the demo URL is substituted with the GitHub repository link.