Show HN: Mint MCP – Generate 3D assets from coding agents

**Mint MCP** is a remote server from the Mint founders that lets coding agents create 3D models, worlds, materials, images, and audio. It is aimed at **Codex**, **Claude Code**, and **Cursor**, and was posted as a Show HN under HN AI Agents. The public entry point is **mcp.mint.gg**, and use requires a Mint subscription; the poster offered to set people up on request.

The product is built as a remote MCP server rather than a local one-off tool. That design fits generation work that can run for a long time: the server owns long-running jobs, preview revisions, partial failures, and file delivery. When a job finishes, Mint returns stable links to the outputs and labels what each file is for so the agent can drop them into the project without guessing paths or roles.

For engineers and builders, the useful part is closing the gap between an agent that can write code and a pipeline that can also produce 3D and media assets in the same session. Instead of leaving the editor to run a separate generator, copy files, and wire them by hand, the agent can request assets, wait through previews and retries, then receive linkable results it can attach to the repo or app.

In the current tooling market, agent hosts already cover code, tests, and some web assets, but 3D and multi-modal generation usually sit outside the agent loop. Mint MCP sits in that gap by exposing generation as MCP tools against a hosted job system. Competition is less about another chat UI and more about other ways to pipe generative 3D and media into agent workflows without custom glue for each editor.

What to watch next is whether agents reliably complete the full loop: request asset, handle previews and partial failures, then place stable links into the project correctly. Subscription access and setup friction will also matter for adoption beyond people who already pay for Mint. The practical test is simple: point Codex, Claude Code, or Cursor at mcp.mint.gg and see whether file delivery and labels are enough for a clean project handoff without manual cleanup.