Submit
Add something to the atlas
Send us a skill, an MCP server or a tool. We read every submission in full and list it with its audit, whatever the verdict. A submission is a GitHub issue, so the whole conversation stays public and permanent.
Cite
Cite this page
ModelsAtlas. “Add something to the atlas”. Retrieved 30 Oct 2026. https://modelsatlas.com/submit
Agent skill
Submit a skill
A public GitHub repository with a SKILL.md, or the directory inside it. We audit the source before listing it.
Open a GitHub issue ↗Opens an issue with the fields filled in.
MCP server
Submit an MCP server
A public repository. We read its exposed tools and classify each one: read, write or destructive.
Open a GitHub issue ↗Opens an issue with the fields filled in.
AI tool
Submit a tool
A product built on AI models, with a public pricing page. We record what its vendor publishes.
Open a GitHub issue ↗Opens an issue with the fields filled in.
What happens to it
- You open an issue
With the link to the repository, or to the directory inside it.
- It is queued ahead of the backlog
Somebody asking beats a crawler finding.
- It is read in full
The instruction text, the code, the dependencies and the manifests, at a fixed commit.
- The findings are published
On its page, whatever they say, each citing the file and line it was found on. A page that says do not install this is a page we publish.
- We keep watching
When the repository changes it is audited again, and the page shows the audits it has had.
What we will not do
- Charge for a listing, a badge or a better verdict.
- Remove a finding because its author asks. If a finding is wrong, say so on the issue with your reasoning and it is re-audited.
- Blend third-party scores into ours. We cite them separately.
- Show a tool's price we have not read on its own pricing page.
Found a wrong price, spec or verdict? Open a correction issue with the page, what is wrong and a source URL for the right value. The thread is public, like every submission. How we check is on the methodology page.