Richard Fu

Posted on Aug 25 Originally published at richardfu.net on Aug 20

Brain, Bridge, Generator: Comparing Unity MCP Options After the Coplay–Aura Acquisition

#ai #gamedev #unity3d #aitooling

My Unity AI setup cost $0 for months, and then the company behind half of it got acquired. That sent me through the Unity MCP alternatives I’d never seriously looked at — and the answer turned out to be less about picking a replacement than about noticing I’d been comparing the wrong three things.

The setup that was working fine

I build a small Unity game on my own. My editor-AI setup arrived by accident rather than design: I started on the free MIT-licensed CoplayDev/unity-mcp bridge, liked it, and after a while moved across to Coplay’s own MCP — the proprietary product from the same shop. That ran for months. It worked genuinely well, and it cost me nothing.

My mental model of what I used it for was simple, and — I’d find out later — wrong. I thought of it as the thing that generated sound effects, plus the occasional texture. Nice to have. Replaceable.

Then I read that Coplay had been acquired, and the two obvious questions arrived together. Is this getting sunset? And is there any actual benefit to moving to the acquirer’s product?

What actually happened

Ramen, the studio behind Aura — an AI assistant for Unreal Engine — acquired Coplay in March 2026, making Aura the first assistant covering both engines. Aura for Unity launched on 14 May 2026. Coplay itself has no announced sunset date and, per the announcement, continues “as a legacy product.”

If you’re happy where you are, nothing needs to change. Unity MCP will also remain free, open source, and MIT-licensed.

— Coplay, What’s next for Coplay

Which is reassuring right up until you notice that the sentence contains a landmine of ambiguity. “Unity MCP” means at least three different things in this ecosystem, and the promise applies to exactly one of them.

Three different things are called Unity MCP

Untangling this was the single most useful ten minutes of the whole exercise, because the free-forever guarantee attaches to the project I’d left, not the one I was using.

Thing	Cost	Asset generation	Fate
CoplayDev/unity-mcp (MIT)	None, ever	None — editor control only	Explicitly guaranteed free and open source; Aura now sponsors it
Coplay MCP (account)	Free tier via a Coplay account	Yes — sfx, music, images, 3D models	Legacy. Running, undeveloped, no end date
Unity’s own MCP (metered)	Bundled on Pro; needs a paid Unity AI seat on Personal	Textures, materials, skyboxes, 3D — no audio	First-party, actively developed

So “the MCP stays free” and “my setup stays free” were two different claims, and I’d been treating them as one.

The reframe: three markets, not one

Every comparison I read mushes these together, which is why the choice looks harder than it is. There are three separable products here, and most people already own one of them outright.

Brain — the agent that reasons and writes the code. If you drive Unity from a coding agent in your terminal or IDE, you already have this and it is the best part of your stack. Nothing below competes with it.

Bridge — the MCP server giving that brain hands inside the editor: read the hierarchy, check compile errors, tail the console, enter play mode, screenshot a scene. This is the part that quietly becomes load-bearing.

Generator — text-to-texture, text-to-mesh, text-to-sound. A completely separate market with its own vendors, and the one where in-editor assistants are acting as resellers.

Once split this way, the acquisition question shrinks. I don’t need a new brain. I don’t need a bundled generator. I need to know that a bridge will exist, and the Unity MCP alternatives turn out to be abundant and free.

Comparing the Unity MCP alternatives

This is a crowded, healthy field — not a three-way vendor choice. Four credible free Unity MCP alternatives, plus the first-party one.

Project	Licence	Tools	What distinguishes it
CoplayDev/unity-mcp	MIT	~47	The default. Largest community by a wide margin, active releases through 2026, sponsored by Aura post-acquisition
IvanMurzak/Unity-MCP	Open	70+	Custom tools via C# attributes. Also runs inside a compiled build, not just the editor
CoderGamester/mcp-unity	Open	—	Mature and broadly client-compatible across the major agents and IDEs
FoxsterDev/xuunity-mcp	MIT	—	Validation-first. Compile checks, test runs, scene assertions, deliberately small mutation surface, no external runtime
Unity official	Seat	—	First-party and durable, but gated behind a Unity AI entitlement unless you’re on Pro or above

Two of these are interesting for non-obvious reasons

Tool counts are the least useful column in that table. The two entries worth actually thinking about earn it on posture, not breadth.

Validation-first is aimed at the failure mode you actually hit. I keep a running file of bridge quirks I’ve been bitten by, and reading it back was instructive: a save-scene tool that silently performed save-as, a script-execution tool that no-ops after a domain reload, a compile check that cheerfully reports a stale build as clean. Almost every entry is a mutation tool lying to me. A bridge whose entire thesis is “small mutation surface, trustworthy compile and test and assert” is pointed directly at that, and nobody markets it that way because “fewer tools” is a terrible tagline.

Custom tools beat vendor tools, and only one option offers them. My real verification loop isn’t generic Unity work — it’s “run the batch harness across a range of inputs”, “replay a recorded line and diff the output against what it produced last time”, “assert the state machine actually reached its terminal state.” I currently reach all of that through fragile generic script-execution calls. Three or four purpose-built tools exposing my own harness would beat any vendor’s eighty-something generic ones, because they’d speak my domain instead of Unity’s. That capability is worth more than every paid feature I looked at, and it’s free.

The paid tier, honestly assessed

I did the work of pricing both, so here it is.

Unity AI gives Personal-plan users a 14-day trial with 1,000 credits, then $10/month for 1,000 credits. Pro, Enterprise and Industry seats include credits and MCP access by default. That last detail matters more than the price: on a Personal seat, the first-party bridge is not a free fallback. There’s an active community thread asking Unity to change this, and another asking for clarity on the policy at all, which tells you how settled it currently feels.

Aura runs a two-week trial with $40 of credit, then Indie, Pro and Ultimate tiers. Its pricing docs contain the detail that decided it for me: when you use its MCP tools from an external IDE, basic inspection is billed to your IDE subscription, but triggering its specialised subagents consumes Aura credits. That’s a fair model — and it means I’d be paying per-invocation for the loop I currently get free, in exchange for an in-editor agent I don’t need, because I already have a brain.

The reseller tell. Coplay’s own comparison page lists 3D model generation “using your provider of choice (e.g. Meshy)” as a premium feature. That’s honest, and it’s also the whole story: the in-editor assistant is a convenience wrapper over generators you can subscribe to directly.

So buy generators from generator companies

If you want generated meshes, the direct vendors are cheaper and better at it than any wrapper: Tripo from around $14–20/month, Meshy at $20, Hyper3D/Rodin with a free generation tier and the strongest reported quality for hero assets. Side-by-side credit pricing is worth a look before committing to any of them.

Two things I’d flag that the marketing doesn’t.

First, generated meshes suit some assets and actively fight others. Anything with radial mechanical symmetry, sharp edges, parts that must stay separable, a pivot that has to sit exactly on an axis, and colliders keyed to geometry — you’ll pay for the generation and then spend the same hours you saved splitting a fused blob apart in Blender. Environment dressing, props, background silhouettes, skyboxes and textures are where this technology currently earns its keep.

Second, audio is the real gap. Unity AI’s generators cover textures, materials, skyboxes and 3D. There is no audio generator at all. If in-editor sfx generation is what you value, no first-party or free-bridge option replaces it, and you’re going direct to an audio vendor whether you like it or not.

On which — a licensing trap worth knowing, since it’s the one place my own “it’s free, it’s fine” reasoning was genuinely broken. ElevenLabs’ free tier is generous on capability: 10,000 credits a month, sound effects included, plenty for a small project. But it grants no commercial licence — free-tier output can’t be used commercially, and non-commercial publishing wants attribution in the title. Starter at $6/month is the cheapest tier that includes commercial rights, and rights persist for whatever you generated while paying. So the correct move isn’t a standing subscription: it’s one paid month covering the session that produces your shipping audio, then back down to free for experimentation. The trigger is “this ships”, not “the credits ran out.”

The option nobody sells you

Worth naming, because it never appears in these comparisons and it may be the most durable choice available: drop the bridge for anything that doesn’t need a live editor.

A Unity batch-mode entry point — -batchmode -executeMethod pointed at your own harness — makes your verification runnable from a plain shell. No MCP server, no vendor, no port collision when you’ve got two editors open, no dependency on anybody’s acquisition. It survives every scenario in this post. The live bridge then stays for the things that genuinely need eyes: screenshots, poking around in play mode, inspecting a hierarchy you can’t picture.

I’d been treating the bridge as the only road into my project from a terminal. It isn’t, and the CLI road doesn’t have a vendor attached.

What I decided

Nothing, immediately — which is a real decision rather than an absence of one. Coplay is up, there’s no sunset date, and I’m a light user of the part that’s actually at risk. Migrating on speculation costs me the accumulated knowledge of one tool’s quirks in exchange for learning another’s.

What changed is that I now hold triggers instead of a diary date.

When the editor plugin breaks on a Unity upgrade → move to one of the free MIT-licensed Unity MCP alternatives. Chosen in advance, no research under pressure.
When generated audio stops being placeholder and ships → one paid month at an audio vendor for the commercial licence, then downgrade.
When props and environment need real volume → subscribe to a 3D generator directly, never through a wrapper.
When I want to spend an hour improving tooling → custom domain tools and a batch-mode entry point, not a bridge swap.

That last one is the actual conclusion. Every one of the Unity MCP alternatives in this post, at best, restores what I already have for free. The only work that makes my setup better is the work no vendor is selling: tools that speak my project’s own vocabulary, and a path into the project that doesn’t route through anybody’s product.

The part that was worth the scare

Legacy products rarely die by announcement. The realistic failure modes are quieter than a sunset notice: an engine upgrade breaks a plugin nobody is patching any more, or an upstream API key silently expires and a generation tool starts returning errors. Waiting for a formal end-of-life date means being surprised by neither of those.

But the acquisition was worth more to me as a forcing function than as a decision. Asked cold what I used the tool for, I said sound effects and the odd texture. Then I actually looked — and the load-bearing dependency was the compile-check, console-tail, play-mode, screenshot loop that I lean on constantly and had stopped noticing precisely because it works. The convenience I’d have named first was almost the only part I could lose without flinching.

That’s the transferable bit, and it has nothing to do with Unity. When a tool in your stack changes hands, the useful response isn’t picking a replacement. It’s inventorying what’s actually load-bearing — because there’s a good chance, as there was for me, that you’d have named the wrong thing.

References

Prices, tiers and licence terms checked August 2026 and certain to drift. This is a decision log, not a benchmark — I compared these on documentation and posture, not by running each bridge through the same project, and I haven’t migrated off anything yet. Corrections welcome.

The post Brain, Bridge, Generator: Comparing Unity MCP Options After the Coplay–Aura Acquisition appeared first on Richard Fu.

DEV Community