Editing Openai/6952a1d0-7e3c-8003-a3f6-55806826a464 (section)

==== ### ====

What it does for you:
* Lets the AI read/write your plugin source, configs, build outputs
* Lets it read BepInEx logs (BepInEx/LogOutput.txt) and iterate fast when something breaks

Most people pair this with a coding assistant in VS Code / Cursor / Claude Desktop. (Filesystem is one of the standard “reference” servers you’ll see in MCP server lists.) GitHub<ref>{{cite web|title=GitHub|url=https://github.com/modelcontextprotocol/servers|publisher=github.com|access-date=2025-12-29}}</ref>

===== What it does for you: =====
* Search your repo history when you broke something
* Pull patterns from other open-source BepInEx plugins
* Generate PR-ready commits/messages and keep changes clean

The Git/GitHub servers are also common “core” MCP picks in curated MCP server lists. GitHub<ref>{{cite web|title=GitHub|url=https://github.com/wong2/awesome-mcp-servers|publisher=github.com|access-date=2025-12-29}}</ref>

===== What it does for you: =====
* Let the AI pull docs and turn them into actionable snippets
* Great for stuff like: BepInEx install quirks, IL2CPP gotchas, config formats

MCP “example servers” commonly include a fetch/web server for exactly this “grab docs → summarize → apply” workflow. Model Context Protocol<ref>{{cite web|title=Model Context Protocol|url=https://modelcontextprotocol.io/examples|publisher=Model Context Protocol|access-date=2025-12-29}}</ref>

(Also: the BepInEx IL2CPP install docs mention first-run generating files + where logs live, which your AI can use to debug “it didn’t load” issues.) BepinEx Docs<ref>{{cite web|title=BepinEx Docs|url=https://docs.bepinex.dev/master/articles/user_guide/installation/unity_il2cpp.html|publisher=BepinEx Docs|access-date=2025-12-29}}</ref>

===== What it does for you: =====
* Store/replay structured debug data: snapshots, events, tick times, desync markers
* Query “when did divergence start?” instead of reading 10k lines of log

There are MCP servers specifically for SQLite workflows. LobeHub<ref>{{cite web|title=LobeHub|url=https://lobehub.com/mcp/simonholm-sqlite-mcp-server|publisher=lobehub.com|access-date=2025-12-29}}</ref>

===== Even though you’re modding a built game, Unity MCP servers can still help because you can: =====
* create a tiny Unity “repro project” to test logic (serialization, interpolation, prediction)
* generate scripts/scenes fast for prototyping systems you’ll later port into the mod

There are multiple Unity MCP server projects that connect the Unity Editor to AI assistants via MCP. GitHub<ref>{{cite web|title=GitHub|url=https://github.com/CoderGamester/mcp-unity|publisher=github.com|access-date=2025-12-29}}</ref>

If your mod includes any networking, this can save you from testing everything inside the actual game executable (aka suffering).