Create
Describe any game. Spawn builds it.

Spawn is a browser place where multiplayer games are made with Savi and played.
spawn.coCrawled Nov 10, 2026
Describe any game. Spawn builds it.
The Spawn desktop app is available for macOS, Windows, and Linux, featuring Discord Rich Presence, auto-updates, and a native game-building window.
A few settings on your computer or phone that make Spawn games run smoother.
Spawn supports user-created mods, including new moves, tools, lights, and maps, which can be added to games.
SpawnJam ran 29 weekly jams: one theme, three days, one game. Every winner is here to play. Games now get featured on the home page through staff picks every week, with no theme or deadline.
Join Spawn and help build the real-time multiplayer engine and AI-native tooling behind multiplayer games made from words.
Maker of Deadwater, Sunday Tips, Buttercup — Cake Studio. 30 games on Spawn.
Maker of Last Light, Onislayer(Nihonggobreaker), Kanji Pop. 4 games on Spawn.
Maker of Swish, Frontier Valley. 2 games on Spawn.
Maker of HABIT SKY MOON. 1 game on Spawn.
Maker of Diablo Unbound, Green Haven. 2 games on Spawn.
Maker of Lady Vesper's Maids, Nine Tails of Fushimi-yama, Starlight Requiem. 40 games on Spawn.
One place to know about Spawn: what it is, the bet behind it, and every engine release.
Spawn lets users collaborate with Savi, an AI game builder, to create and instantly play games together in a browser.
Spawn deploys AI agents that automatically push code updates to its live games on an hourly basis. The company prioritizes rapid recovery over preventive measures, demonstrating a hands-off approach to live operations. A documented failure highlights the risks of this strategy, serving as a cautionary example for real-time systems.
Spawn reimagines company operations by abandoning traditional startup structures and AI-driven workflows. Unlike most firms optimizing existing processes with AI, Spawn integrates AI into its core model, reshaping roles and outputs entirely. The company exemplifies a radical departure from conventional organizational frameworks, prioritizing AI-native design over incremental improvements.
Startup **Spawn** uses a swarm of AI agents powered by the **Fable model** to dynamically generate and update its massively multiplayer online game. Players interact with an embedded AI assistant named **Savvy** via text commands to modify the game world, such as constructing structures or adding creatures. The AI agents autonomously edit live game code and deploy updates to players hourly, though they occasionally introduce bugs. Token costs for AI model usage exceed **$3 million monthly**, far surpassing the company’s total salary expenses.
Spawn’s AI agents automatically resolve user errors without human intervention. The company’s ten-person team invests 30 times its total payroll on AI token costs to maintain these autonomous systems. Developers rely on the agents to handle issues seamlessly, reducing manual troubleshooting.
Spawn founder Jacob Sansbury asserts a team of ten developers deployed **77,000 code changes** over **30 days** using an AI agent fleet, consuming **$3.1 million in token costs**. The claim contrasts claimed versus verified output, with no traditional pull requests required. The post distinguishes between what the system *claims* to achieve and what is independently *verified* in production.
Spawn.co attempts to generate music games but encounters technical issues. The platform struggles with initialization errors and poor starting conditions, frequently causing crashes during gameplay. Developers face instability when testing generated music game prototypes.