Skip to content

MCP for Modrinth #6457

@Pulsar-Programmer

Description

@Pulsar-Programmer

Please confirm the following.

What parts of Modrinth is your feature request related too?

No response

Is your suggested feature related to a problem? Please describe.

I would love an MCP for Modrinth that could create collections, modpacks, and organize things. It would make things a lot quicker for the user.
For example, I wanted to look through my saved mods in a collection, filter them into Hoplite-compatible mods, and then download them all from the collection.

Describe the solution you'd like

Model Context Protocol (MCP) server for Modrinth that exposes the Modrinth API to AI assistants (Claude, etc.) as a set of tools.
At minimum it would wrap the public Modrinth API to support things like:

Search projects: query mods, modpacks, resource packs, shaders, plugins, and datapacks with filters for loader (Fabric, Forge, Quilt, NeoForge), Minecraft version, category, and project type.
Get project details: fetch a project's description, supported versions, license, gallery, and stats by slug or ID.
List/get version: retrieve a project's version list and individual version metadata, including download URLs, dependencies, and which game/loader versions each supports.
Resolve dependencie: given a project or version, return its required/optional dependencies so a model can help assemble a compatible modlist.
User/team lookup: fetch author profiles and the projects they maintain.

The goal is for someone to ask an assistant something like "find me a performance-focused Fabric mod for 1.21 and list its dependencies," and have the model call the MCP tools to answer with live, accurate data.

It could be built on the existing @modrinth/api-client (official TS client) rather than reimplementing HTTP calls, and would set a compliant User-Agent per Modrinth's API guidelines. Authenticated/write endpoints (creating versions, editing projects) could come in a later phase behind a personal access token.

Describe alternatives you've considered

I suppose one could make their own and hook it up to the API, but a native solution might be cleaner.

Additional context

No response

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels
    No fields configured for Feature.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions