CloudSprite MCPサーバー
コミット 2026年10月25日Allows AI clients to search CloudSprite datasets, documentation, and notebooks while managing organizational scope and submitting feedback.
別名 cloudsprite-io/cloudsprite-plugin, cloudsprite-plugin
- スター
- 1
- ツール
- —
- ダウンロード数(30日間)
- —
- 最新バージョン
- 0.1.3
- 最終コミット
- 2026年10月25日
- ライセンス
- MIT
CloudSprite MCPとは
CloudSpriteはMCPサーバーです。 Allows AI clients to search CloudSprite datasets, documentation, and notebooks while managing organizational scope and submitting feedback.
プロジェクトのREADMEから生成された概要
CloudSprite MCP server について
CloudSprite provides a remote MCP connection that allows AI clients to interact with the CloudSprite platform. It is designed for developers using tools like Claude Code, Grok Build, Cursor, GitHub Copilot, and VS Code. The server operates over streamable HTTP and uses OAuth for authentication, meaning no API keys or secrets are stored within the plugin itself. Users sign in using their existing CloudSprite browser accounts to establish a secure session.
This implementation is currently read-only, focusing on information retrieval and user feedback. It enables models to explore organizational context, including teams and projects, and search through datasets, notebooks, and scripts. Developers can also query product documentation and the SDK catalog. While it does not modify data, it includes a specific tool for submitting bug reports or feature requests directly to the CloudSprite team via the submit_feedback command.
できること
- 1Explore organizational structureAsk the AI to list your available organizations, teams, or specific projects to establish context.
- 2Submit platform feedbackTell the AI to report a bug or suggest a new feature for the CloudSprite service.
プロジェクトの README から作成されています。すべてのユースケースは、その引用と照らし合わせて確認されています。
CloudSprite MCPのインストール方法
フォーマット検証日 2026年11月10日Cursor に追加Gooseに追加Claude CodeにCloudSprite MCPをインストール
$ claude mcp add --transport http cloudsprite-plugin https://api.cloudsprite.io/mcpCursorにCloudSprite MCPをインストール
~/.cursor/mcp.json
{
"mcpServers": {
"cloudsprite-plugin": {
"url": "https://api.cloudsprite.io/mcp"
}
}
}VS CodeにCloudSprite MCPをインストール
.vscode/mcp.json
{
"servers": {
"cloudsprite-plugin": {
"type": "http",
"url": "https://api.cloudsprite.io/mcp"
}
}
}VS Code CLIにCloudSprite MCPをインストール
$ code --add-mcp '{"name":"cloudsprite-plugin","type":"http","url":"https://api.cloudsprite.io/mcp"}'CodexにCloudSprite MCPをインストール
$ codex mcp add cloudsprite-plugin --url https://api.cloudsprite.io/mcpCodex config.tomlにCloudSprite MCPをインストール
~/.codex/config.toml
[mcp_servers.cloudsprite-plugin] url = "https://api.cloudsprite.io/mcp"
Gemini CLIにCloudSprite MCPをインストール
$ gemini mcp add -s user --transport http cloudsprite-plugin https://api.cloudsprite.io/mcpGemini settings.jsonにCloudSprite MCPをインストール
~/.gemini/settings.json
{
"mcpServers": {
"cloudsprite-plugin": {
"url": "https://api.cloudsprite.io/mcp",
"type": "http"
}
}
}自動チェック
実行日 2026年11月14日完全なレポート →まだ検証されていません。サーバーには80点のスコア、重大な問題がないこと、および検証に必要な十分なデータが必要です。 チェックの仕組み · 点数を上げるための改善点
- ソースリポジトリがリンクされており、アクセス可能cloudsprite-io/cloudsprite-plugin is public and reachablegithub.com/cloudsprite-io/cloudsprite-plugin ↗合格
- パッケージがこのリポジトリをソースとして指定未チェック
- ビルドの来歴 (Sigstore 証明)未チェック
- オープンソースライセンス (OSI)MITspdx.org/licenses/MIT.html ↗合格
- 過去1年以内にコミットまたはリリースありLast commit or release 2026-10-25github.com/cloudsprite-io/cloudsprite-plugin ↗合格
- OpenSSF Scorecard未チェック
- パッケージ名が有名なパッケージに酷似していない未チェック
- パッケージおよびその依存関係に重大な勧告なし未チェック
- モデルへの隠れた指示がない0 tool descriptions and the README read; no override phrases, credential paths or hidden characters合格
- 破壊的なツールがラベル付けされている未チェック
- リモートエンドポイントが MCP ハンドシェイクに応答HTTPS endpoint answered 401 (sign-in required)合格
README
GitHubで見るCloudSprite plugin
Public plugin for CloudSprite. One repo, one plugin slug (cloudsprite), for Claude Code, Grok Build, and Agent Plugins clients (Cursor, GitHub Copilot, VS Code, Codex, and others).
The plugin talks to production CloudSprite over remote MCP. There are no API keys, GitLab tokens, Linear tokens, or other secrets in this repository. OAuth tokens stay in the client.
Plugin name is cloudsprite and is immutable once a marketplace lists it. The GitHub org is cloudsprite-io because cloudsprite was already taken on GitHub.
Connect your AI to CloudSprite
Full per-client steps: Connect your AI to CloudSprite.
There is no API key in this repo and none in chat. Sign in with the same CloudSprite account you use in the browser.
| What | Value |
|---|---|
| MCP URL | https://api.cloudsprite.io/mcp |
| Transport | Streamable HTTP |
| OAuth discovery | https://api.cloudsprite.io/.well-known/oauth-authorization-server |
| Plugin slug | cloudsprite (fixed) |
How do I connect Claude Code to CloudSprite?
/plugin marketplace add cloudsprite-io/cloudsprite-plugin
/plugin install cloudsprite@cloudsprite
Enable the cloudsprite plugin if it is not already on. On first tool use, complete the browser sign-in.
How do I connect Cursor, GitHub Copilot, VS Code, or Codex to CloudSprite?
This repo is an Agent Plugins 1.0 package: root plugin.json, mcp.json, and skills/.
- Clone this repository (or add
cloudsprite-io/cloudsprite-pluginas the plugin source if your client accepts a git URL). - Load it as an Agent Plugin. In Cursor, a local checkout under
~/.cursor/plugins/local/cloudspriteis enough. - Enable the plugin and complete OAuth when the client prompts.
If the client cannot load Agent Plugins, add a remote MCP server instead (same URL and OAuth as any MCP client).
How do I connect Grok to CloudSprite?
grok plugin marketplace add cloudsprite-io/cloudsprite-plugin
grok plugin install cloudsprite --trust
Official listing on xAI’s marketplace is separate and not required for this install. Maintainer catalog notes: DIRECTORY.md.
How do I connect any MCP client to CloudSprite?
| Setting | Value |
|---|---|
| Server URL | https://api.cloudsprite.io/mcp |
| Transport | Streamable HTTP |
| OAuth discovery | https://api.cloudsprite.io/.well-known/oauth-authorization-server |
The client must discover authorization, token, and JWKS URLs from that metadata. Do not hard-code those hosts, and do not put a client secret or API key in the config.
{
"mcpServers": {
"cloudsprite": {
"type": "streamable-http",
"url": "https://api.cloudsprite.io/mcp"
}
}
}
Some clients use "type": "http" or a bare "url" field for the same remote server.
How do I connect ChatGPT to CloudSprite?
ChatGPT connectors are coming soon. Do not add CloudSprite as a ChatGPT custom connector yet — that OAuth redirect is not registered. Use Claude Code, Cursor, Copilot, Codex, Grok, or a generic MCP client today.
What it does
After you sign in and set scope, the model can:
- Read org / team / project context (
whoami/get_scope,list_orgs,list_teams,list_projects,set_scope) - Search datasets, traces (summaries only), notebooks, and scripts
- Search CloudSprite product docs and the SDK catalog (
search_knowledge,search_sdk) - File a bug or feature request (
/feedback→ MCPsubmit_feedback)
This release is read-only plus feedback. It does not mutate datasets, tags, notebooks, or scripts.
MCP URL
Production (the only URL customers use):
https://api.cloudsprite.io/mcp
Transport: MCP Streamable HTTP. No Authorization header is stored in the plugin. The client discovers OAuth from the MCP origin and holds the tokens. There is no environment picker in the plugin UI.
A 401 from /mcp before OAuth is live is expected. The plugin should fail with a clear auth error, not hang.
Claude Code honors CLOUDSPRITE_MCP_URL if the client process has it set (${CLOUDSPRITE_MCP_URL:-https://api.cloudsprite.io/mcp} in .mcp.json). The GUI app reads ~/.claude/settings.json "env", not your shell profile. Portable mcp.json is a literal production URL (no interpolation). Do not commit a .env here.
Skills
| Skill | Slash | Purpose |
|-|-|
| feedback | /feedback | Confirm text, then call MCP submit_feedback |
| platform-howtos | /platform-howtos | Sign-in, scope, and how to use CloudSprite from the assistant |
| datasets | /datasets | Find datasets, parameters, tags, and notebooks (read) |
| mixed-mode-analysis | /mixed-mode-analysis | Differential / common-mode S-parameters from 2-port pairs |
| waveform-correlation | /waveform-correlation | Pairwise Pearson QC across repeated traces |
Network endpoints
The plugin package does not open sockets itself. Clients that load it will contact:
| URL | Why |
|---|---|
https://api.cloudsprite.io/mcp |
MCP Streamable HTTP (tools) |
https://api.cloudsprite.io/.well-known/oauth-authorization-server |
OAuth 2.1 authorization-server metadata |
https://api.cloudsprite.io/.well-known/oauth-protected-resource |
OAuth protected-resource metadata (if advertised) |
| Authorization, token, revocation, and JWKS URLs from that metadata | Cognito Hosted UI PKCE. Hosts are not hardcoded here; they come from discovery. |
Installing the plugin also clones this GitHub repository (https://github.com/cloudsprite-io/cloudsprite-plugin). Tagged releases are vMAJOR.MINOR.PATCH matching plugin.json version.
The plugin does not call GitLab, Linear, HubSpot, or any issue tracker. Feedback is filed by the CloudSprite API after submit_feedback.
If CLOUDSPRITE_MCP_URL is set, discovery and token endpoints follow that host instead of api.cloudsprite.io.
Support
- Product: https://cloudsprite.io
- Docs: https://docs.cloudsprite.io
- Email: support@cloudsprite.io
- Privacy: privacy@cloudsprite.io
Releases
Git tags are v + the version in plugin.json (currently 0.1.3). Maintainers publish with scripts/publish.sh after scripts/audit.sh passes. Republishing the same tag is a no-op if HEAD still matches; a different tree at the same version fails.
License
MIT. Copyright © 2026 Brushfield Ventures LLC d/b/a CloudSprite.
Product use of CloudSprite remains under the CloudSprite Terms of Service and Privacy Policy.
CloudSprite MCPのよくある質問
CloudSprite MCPをClaude Codeに追加するには
ターミナルでclaude mcp add --transport http cloudsprite-plugin https://api.cloudsprite.io/mcpを実行します。すべてのプロジェクトで利用するには--scope userを追加します。
CloudSprite MCPをCursorに追加するには
~/.cursor/mcp.jsonを開き、上記のインストールセクションに示すCursor設定を使ってmcpServersの下にサーバーを追加します。
Does this plugin require an API key?
No, the plugin does not use API keys or secrets; it uses OAuth tokens that stay within the client.
Can the AI modify my datasets or notebooks?
No, this release is read-only for datasets, tags, notebooks, and scripts, though you can file feedback.
How do I connect Cursor or VS Code to CloudSprite?
Clone the repository and load it as an Agent Plugin, then enable the plugin and complete the OAuth prompt.