Two People, 26 Releases, 3,400 Developers: The UE5 AI Copilot 2.0 Turning the Engine Into an 'AI Game Studio'
On October 2, European indie studio BlueprintsLab released Ultimate Engine CoPilot V2 — the most widely adopted AI plugin for Unreal Engine 5, with 1,655 native tools across 37 extensions. It doesn't tell you where to click; it builds Blueprints, writes C++, runs multiplayer play tests, and checks its own work inside the editor. The whole team: two people.
What happened: the most popular AI plugin for Unreal Engine rebuilt from the ground up
On October 2, European indie studio BlueprintsLab released Ultimate Engine CoPilot V2 — the most widely adopted AI plugin for Unreal Engine 5. This is a ground-up rebuild: 1,655 native tools across 37 extensions. The AI is no longer a chatbox telling you where to click; it does the work inside the editor — building Blueprints, writing C++, doing VFX, laying out levels, animating, cutting cinematics — then compiles, plays the level, and checks its own work.
It serves 3,400+ developers, holds 250+ reviews on Epic's Fab marketplace with two official features, and has run 9.4 million AI actions inside real projects, generating 203,000+ assets and 32,000+ C++ files. The team behind all of it: two people — founder Roly worked solo for the first four months.
V2's core: not a chatbox, a self-inspecting construction crew
V2's architecture is called Crew: a Planner, a Worker, and a Reviewer, each on its own model. The Worker builds step by step, the Reviewer checks against success criteria, and nothing counts until it passes. The hardcore capabilities:
Design docs to build plan in 2 minutes: drop in nine game design documents and get a 38-task build plan — timed at 2 minutes in a recorded demo. One-click Blueprint to C++: converts a Blueprint into a new C++ class, compiles it, fixes its own compile errors — under 4 minutes on a real project. It plays your game: automated play tests run the level after every change; multiplayer tests spin up a listen or dedicated server with up to 8 clients and compare what each one sees. It finds your frame-rate killers: one-click Unreal Insights traces, with the AI naming the Blueprint behind the slowest frames.
Radically open: models, keys, agents — all yours
V2 locks you into no model: use BlueprintsLab's hosted BpLabs Premium (for hard, long builds), bring your own keys (OpenAI, Claude, Gemini, DeepSeek, OpenRouter), or local models (Ollama, LM Studio) — studios under NDA can keep every byte on their own machines. Even bolder: Claude Code, OpenAI Codex, GitHub Copilot, and Gemini CLI run directly inside the Unreal editor, with a built-in MCP server for external clients.
Pricing: around $200 per seat one-time; Indie at $25/month (with a token pool); Studio at $60/month. V1 owners upgrade free. Supports UE 5.7/5.8 on Windows and Mac, with Linux on the way.
Our take: this is what vibe coding's "iPhone moment" in games might look like
The most remarkable part of this story isn't the feature list — it's how it's built: 26 releases in 14 months, roughly one every two weeks; after V2.0.0 launched on September 30, versions 2.0.3, 2.0.5, and 2.0.6 shipped within 48 hours — all from developer feedback on Discord. Founder Roly's own words: "No ticket queue, no roadmap committee. You tell us what slows you down, and it's fixed in the next update."
While Aura, Ludus AI, and NeoStack — teams with big funding and dozens of engineers — plus Epic's own built-in MCP integration all chase it, a two-person team holds first place by staying closest to users. That's the ultimate validation of the vibe coding spirit: in the AI era, what caps a product is no longer team size, but iteration speed and a nose for real pain. For indie game developers, "one person is a game studio" is turning from slogan into a shopping list.
Sources
Related articles
The Zephos team planted 16 launch-killer bugs in Notely, an agent-built Next.js + Supabase + Stripe notes app — two payment-related: unsigned webhooks accepted, pro granted from a self-declared client_reference_id. This guide turns those traps into a playbook: webhook signature verification, a server-side single source of truth, the subscription state machine, test clocks, and a launch checklist. Money logic must be hand-written or audited line by line.
Reported by InfoQ on October 3: a GPT-4 coding agent given CVE descriptions successfully exploited 87% of 15 test vulnerabilities, versus 7% without descriptions. rclone's author received 40+ security disclosures in a single month — more than the project's previous decade combined; QEMU has shortened its embargo period. The vulnerability disclosure timeline is collapsing under agent speed.
On October 7, OutSystems announced Agent Experience is generally available: its low-code platform is now open to any AI coding agent — Claude Code, Cursor, Codex, Kiro — with agents working at the design level, the platform generating code deterministically, and governance built in. This is the "vibe coding goes enterprise" playbook: taming shadow AI with a compliant path. But the 74% rework figure is vendor-survey data — discount it. The real bill is the hidden cost of platform lock-in.