Co-op Anything Lets AI Agents Add Multiplayer to Any Single-Player Game

Co-op Anything, an open-source project for building multiplayer game mods, gives AI coding agents a repeatable workflow for adding co-op to single-player games. The repository combines research guidance, engine-specific techniques, synchronization tests, packaging instructions, and a Model Context Protocol server for tools such as Claude Code, Codex, Cursor, and Gemini CLI.

The project does not provide one universal mod. A developer points an agent at a locally installed game, then the agent searches for prior work, identifies the engine, inspects available code or binaries, chooses an implementation strategy, and tests the resulting mod. Successful title-specific projects can package installation and Steam-based connections for players, although the development process remains hands-on. At the time of publication, the repository reported 1.1k GitHub stars.

Agents get two routes

After a developer names a game, the playbook directs the coding agent down one of two paths:

• Repair existing work: The agent searches repositories, forums, issue trackers, and available community discussions for an unfinished or abandoned mod. It audits the code, reproduces reported failures, adds regression tests, and publishes fixes as a separate package.
• Start from scratch: The agent identifies the engine and extension points, adds another player, then synchronizes movement, enemies, world state, quests, inventories, and saves in stages.

Each phase maps to a reusable skill stored under .agents/skills or Claude Code’s .claude/skills directory. The included skills cover research, revival work, reverse engineering, greenfield development, networking, synchronization testing, and quality assurance.

An MCP front door and a test lab

The Python-based MCP server exposes project functions to compatible coding agents through Model Context Protocol, a standard for connecting models to external tools and data. An agent can search the co-op directory, inspect a game installation to identify its engine, load the relevant playbook, and validate a proposed contribution.