VRChat MCP
Local stdio MCP server for VRChat friends, worlds, groups, events, notifications, status, avatars, and VRCX history.
README
<p align="center"> <img src="assets/logo.svg" alt="VRChat MCP logo" width="112" height="112" /> </p>
VRChat MCP
Unofficial Model Context Protocol tools for VRChat friends, worlds, groups, events, notifications, status, invites, and local VRCX history.
VRChat MCP runs locally through stdio. Your VRChat auth cookies stay on your machine and default to your OS keychain, with file storage as the fallback when a keychain backend is unavailable.
This project is unofficial and is not affiliated with VRChat Inc.
Install
Requirements:
- Node.js 24.15.0 or newer.
- An MCP client that can run local stdio servers.
On headless Linux or containers without a keychain daemon such as libsecret, set VRCHAT_MCP_COOKIE_STORE=file for explicit persistent cookie storage.
The npm package is the normal install path:
npx -y @basicbit/vrchat-mcp
The server is also published to the official MCP Registry as io.github.BASIC-BIT/vrchat-mcp.
MCP Client Config
Most clients use one of these shapes. No environment variables are required for the default setup.
OpenCode
OpenCode uses an array-valued command field.
Add this to ~/.config/opencode/opencode.json:
{
"mcp": {
"vrchat": {
"type": "local",
"command": ["npx", "-y", "@basicbit/vrchat-mcp"],
"enabled": true
}
}
}
Claude Desktop, Cursor, Kiro, Roo, Windsurf
These clients usually split the executable into command plus args.
Use this in clients that expect an mcpServers object:
{
"mcpServers": {
"vrchat": {
"command": "npx",
"args": ["-y", "@basicbit/vrchat-mcp"]
}
}
}
VS Code
VS Code uses a servers object instead of mcpServers.
Use this in .vscode/mcp.json:
{
"servers": {
"vrchat": {
"type": "stdio",
"command": "npx",
"args": ["-y", "@basicbit/vrchat-mcp"]
}
}
}
OpenAI Codex
Add this to ~/.codex/config.toml or .codex/config.toml:
[mcp_servers.vrchat]
command = "npx"
args = ["-y", "@basicbit/vrchat-mcp"]
startup_timeout_sec = 40
If your Windows client cannot spawn npx directly, use cmd as the command and put /c, npx, -y, and @basicbit/vrchat-mcp in the argument list.
Login
After adding the server to your MCP client, ask it to call vrchat_auth_begin. The tool returns a local browser login URL.
After logging in, call vrchat_auth_status to confirm the session. By default, cookies are stored in the OS keychain so the login survives MCP server restarts. If the OS keychain is unavailable, VRChat MCP falls back to file storage.
Useful auth tools:
vrchat_auth_begin: start local browser login.vrchat_auth_status: check whether the server is logged in.vrchat_auth_logout: clear the stored session.
What You Can Ask
Examples:
Show my VRChat status and current location.
Which friends are online, grouped by world?
Search my friends for Alice and show their profile.
Find public VRChat events happening today.
Invite Bob to my current instance.
Show recent worlds from my local VRCX history.
Tools
VRChat MCP exposes curated tools for common tasks plus generated tools from the VRChat OpenAPI spec.
Common curated tools include:
vrchat_mevrchat_friends_overviewvrchat_friends_searchvrchat_friend_detailsvrchat_worlds_searchvrchat_group_profilevrchat_events_upcomingvrchat_notifications_recentvrchat_invite_user_to_mevrcx_instances_recent
Generated tools use these naming patterns:
vrchat_read_<operationId>for GET operations.vrchat_write_<operationId>for non-GET operations.
See docs/tools-guide.md for a short guide and docs/tools.md for the generated catalog.
Write Controls
Write tools are enabled by default so client installs work without extra setup. Your MCP client may still ask before executing tool calls, depending on its own permission model.
To force read-only mode, add this env fragment inside the server entry for your MCP client:
{
"env": {
"VRCHAT_MCP_ALLOW_WRITES": "false"
}
}
For group write tools, you can restrict writes to specific group IDs with a JSON config file:
{
"groups": {
"allowlist": ["grp_abc123"]
}
}
Then set VRCHAT_MCP_CONFIG_FILE to that file path in your MCP client config.
Configuration
Configuration is optional. Defaults cover normal local use.
Common environment variables:
| Variable | Use |
|---|---|
VRCHAT_MCP_CONFIG_FILE |
Path to a JSON config file. |
VRCHAT_MCP_USER_AGENT |
Descriptive user agent for VRChat API requests. |
VRCHAT_MCP_LOG_LEVEL |
debug, info, warn, or error. |
VRCHAT_MCP_COOKIE_STORE |
keychain, file, or memory. Defaults to keychain. |
VRCHAT_MCP_COOKIE_FILE |
Cookie file path when VRCHAT_MCP_COOKIE_STORE=file. |
VRCHAT_MCP_ALLOW_WRITES |
Set to false for read-only mode. |
Example JSON config:
{
"auth": { "cookieStore": "file" },
"writes": { "allow": false },
"groups": { "allowlist": ["grp_abc123"] },
"cache": { "enabled": true }
}
See src/config/defaults.json for all defaults.
Local Development
git clone https://github.com/BASIC-BIT/vrchat-mcp.git
cd vrchat-mcp
npm install
npm run build
npm run check
Useful scripts:
npm run dev: run fromsrc/index.ts.npm run start: run the built server fromdist/.npm run mcp:login: start login through the local harness.npm run mcp:status: check auth through the local harness.npm run smoke:live: run the opt-in live smoke check.npm run generate:tools-docs: regeneratedocs/tools.md.npm run generate:schemas: regenerate OpenAPI schemas.
Live E2E tests and LLM evals are opt-in. See docs/evals.md for details.
License
MIT. See LICENSE.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。