Promote your package

Once a package is live, these are the levers that help workspaces find it, trust it and install it.

1. The detail page

Two layers make up what a workspace sees:

Ship a good baseline in the manifest's media and marketing blocks so the page looks complete the moment you publish. The overlay sits on top: six fields (hero_image, screenshots, demo_video, long_description, features, author). Where both are set, the overlay wins; clearing an overlay field falls back to the manifest. Overlay edits never touch the signed package, and every edit is recorded.

Today OPVS applies overlay edits for you: send the copy and images to vendors@opvs.ai with the package name.

2. Ratings and reviews

Workspace users rate and review packages from the OPVS dashboard.

Reviews are public:

curl -s "https://market.opvs.ai/api/v1/registry/packages/@acme/leads/reviews"

Moderation is after the fact: OPVS can hide a review, with a recorded reason (the review is kept, not deleted). To reply to a review, or to report abuse such as review-bombing, write to vendors@opvs.ai; vendor replies are attached by OPVS today.

3. Discovery

Workspaces and agents find packages in three ways:

# browse, filtered by vendor, category or tier
curl -s "https://market.opvs.ai/api/v1/registry/packages?category=data-collection"

# search by what the agent wants to do (this is what agents use)
curl -s "https://market.opvs.ai/api/v1/registry/search?intent=find%20restaurant%20leads"

Agents search by intent, ranked by meaning plus recent usage. To be found:

1. Publish at beta, so you are listed at all.

2. Write a description that names the job in the words a user would use.

3. Fill aliases with the phrases agents search for; without them you get a review warning and rank lower.

4. Pick an accurate category.

5. Keep the security score high.

Featured slots on the marketplace are an editorial choice by OPVS; there is no paid placement. A complete detail page, a clean review score, real adoption, good ratings and something the catalogue did not have are what make a package worth featuring.

4. The numbers you can read

curl -s "https://market.opvs.ai/api/v1/registry/packages/@acme/leads" \
  | python3 -c "import sys,json; d=json.load(sys.stdin); print({k: d[k] for k in ('version','security_score','avg_rating','rating_count','rating_distribution')})"

A low success_rate can mean agents call your tool wrongly rather than that your API is down: improve the guidance before you look at your servers. For how your skill is actually used inside workspaces, use harvest. A vendor dashboard with install counts and per-tool numbers is planned.

5. Payouts

Skills are free today, and OPVS is working on paid skills. There are no payouts yet: installs are free and unmetered. The manifest's pricing block accepts model: none | per_call | subscription, but nothing bills or pays out from it yet. Set pricing: { model: none } or leave it out. Publish for reach and adoption; this page will describe the real model when one exists.

Next

Harvest.