Agents shop here too.
The catalog speaks MCP. An agent asks for a rusty barrel under 5k tris and gets the right oil drum with everything behind it, even though no filename says barrel.
Get my early-access seat

An agent can block out a level; it can't go digging through your network drive. Give it the library over MCP and it asks for a rusty barrel under 5k tris, gets a real oil drum with its dependencies, and places it, instead of inventing a path and hoping.
Any MCP client connects: Claude, Cursor, or your in-house tooling. The index answers in assets, with tri counts, packs, and dependencies attached, so the agent's next call can be a pull rather than another question.
Access is tenant-scoped and authenticated, and it's read-only until you switch on write. An agent can never ship something your library doesn't hold.
Fair questions.
- Is our library feeding someone's model?
- No. Nothing you store trains anything. An agent sees search results from your workspace and nothing else.
- What happens when an agent runs wild?
- Agent calls draw from a plan allowance and pause when it runs out. A runaway loop hits a ceiling, and searching from the apps stays unlimited either way.
- How much wiring does this take?
- It's an endpoint and a key, scoped to your workspace. Point Claude, Cursor, or your own MCP client at it and the tools show up.
Your agents block out the level.
Your library supplies it.
An agent dressing a set at 2 a.m. pulls from the same catalog your artists use at 2 p.m. Same search, same assets.
Try it on your real library.
- A hosted seat for your studio when the next early-access batch opens.
- Your first pack searchable by sight in about four minutes.
- Works with Unreal Engine 5.3 and up.
In private testing with working studios now. Seats open in small batches; the list is the queue.
Prophouse