Weva decides who may use which AI skill

and runs it for them. Everyone in Weva can see what exists; only the people an owner says yes to, and the circles they share it with, can use it, and they need no model subscription of their own.

Sharing is caring

How it works

Open to look at, closed to use, and it runs without a key of your own.

  1. Discover

    Every Weva-wide skill has a card: what it does and who made it. Every member can read those cards. Only the people its owner says yes to can open the skill itself.

  2. Ask

    Found something useful? Ask the owner. If they say yes, you can use that skill. You join none of their circles, and they can take it back whenever they want.

  3. Use

    Once they say yes, the skill works in your own tools. Weva provides the model access it needs, so there is no model key to obtain and no provider account to open.


Three situations this is for

Concrete ones, because that is how you will know whether it is yours.

revoked at handover

An agency and its client

You built a set of agent skills for a client. They should have them while the project runs, and not afterwards. Create a circle for the people on the project, share the skills into it, remove them at handover, and show exactly which versions they received.

runs anyway

Someone with no API key

The person you are sharing with has ChatGPT and nothing else - no OpenAI billing, no OpenRouter account, no appetite for either. They open the skill and it works, because Weva makes the model calls and counts them against a monthly allowance.

contribute

A team improving one skill

Share it into the circle with Contribute access and anyone there can propose a better version. The owner accepts it, and the person who wrote it keeps the credit. One history, one author per version.


It arrives where you already work

Two ways in, depending on how your tools are shaped.

In your editor

A packaged skill unpacks straight into Claude Code or Cursor, where skills already live as files.

In your chat

A connector puts them inside Claude on the web and in Desktop, Claude Code and Cursor, and ChatGPT on the web, so you can ask for one by name instead of finding a file.


Why it works without your own API key

A shared skill is no use if the person receiving it has to open a billing account first. So Weva holds the model access and makes the calls on your behalf, starting with a small, fast model and moving to a stronger one only when the answer needs it. Most requests never leave the cheapest option.

Each member gets a monthly allowance. A run is priced before it starts and declined if it would go over, so nobody is surprised by a bill - including the person who pays for Weva.


How it is put together

One service in the middle, your own tools at both ends. Weva decides who reaches what, and makes the model calls - it never runs a skill.

Three bands. Left: a skill folder, packaged and checked for
                portability. Middle: Weva holds a directory of cards every
                member can read, the access an owner gives, to one person or
                to a circle, a gateway that makes the model calls, and an
                audit trail. Right: the packaged skill unpacks into Claude
                Code or Cursor as files, a connector puts it inside Claude,
                ChatGPT on the web, Claude Code and Cursor,
                and your own tool runs the skill and asks the gateway for a
                model.
Publishing on the left, Weva in the middle, the people you approved on the right. The gateway is where the model calls happen. The dashed lines are the audit trail.

What Weva is not

Saying this plainly is what makes the rest believable.

  • Not a public marketplace. Membership is by invitation, and nothing is listed for people outside Weva to browse.
  • Not a place that runs other people's code. Weva makes the model calls a skill asks for, and nothing else executes on our side.
  • Not a safety authority. Weva does not review or vouch for what you share, and says so where it matters.
  • Not a reseller of model tokens. What is worth paying for is the routing that keeps a good answer cheap, not the tokens underneath it.
  • Not a replacement for git. If everyone you share with is comfortable with a private repository, use one.

The first skill shared through Weva

BlenderRender turns a sentence into a rendered 3D picture. Read the walkthrough to see what receiving a skill looks like today - and what the registry has to automate away.

Read the walkthrough

The introduction, as a deck

Eight slides on why Weva exists, what it does and does not do, how it is put together, and what it looks like in your week. It comes in the theme you are reading in. PowerPoint, 84 KB.

Download the deck

Tell me if this is your problem

If this is a problem you have, leave an address and I will let you know when there is something to try.

Your address is stored on Weva's own server, used only to tell you when Weva opens, and never passed on. Ask any time and it is deleted.