AI Developers

Hi everyone,I’m testing the Box MCP Server with ChatGPT and have run into an issue specifically with write/update operations.Read operations work normally. I can search folders, list folder contents, read file metadata, and retrieve file contents without any problem.However, when I try to update an existing file, the MCP tool schema shows that tools such as upload_file_version and get_upload_url are available, but when ChatGPT actually invokes them, the runtime returns:Tool upload_file_version not foundandTool get_upload_url not foundThe affected file is a normal Markdown (.md) file in Box, and the authenticated user appears to have upload/edit permissions on the folder.So the situation seems to be: Box MCP connection/authentication works Read tools work The write/upload tools appear in the exposed tool list/schema But those tools are not available at runtime Has anyone encountered this behavior with the Box MCP Server?Is there an additional Box MCP configuration, permission, fea

Can you have 2 Box accounts and Box Drive with different Email addresses on the one laptop?

 We are building a web-based LLM platform (similar to Claude and ChatGPT) and are planning to integrate with Box via MCP (Model Context Protocol). We have noted that Box hosts a remote MCP server at https://mcp.box.com and would like to proceed with the integration. Before doing so, we have a few prerequisite questions:1. MCP Integration Prerequisites: What are the prerequisites to integrate with Box's MCP server? Is registration or access approval required before connecting?2. Redirect URL / Whitelist Requirement: For OAuth-based MCP authentication, do we need to register or whitelist any redirect URLs on your end? If yes, where do we submit or register them?3. Marketplace / App Listing: We noticed your integrations page at Box MCP Server - https://developer.box.com/guides/box-mcp#platform-setup-guides, where tools like Claude, ChatGPT, Microsoft Copilot, and Cursor are listed as MCP-compatible integrations. Since our product is a similar LLM platform, we would like to explore getting

お世話になっております。 現在、Boxが提供するリモートMCP（Model Context Protocol）サーバー（ https://mcp.box.com ）を利用し、外部LLMとの統合テストを行っております。接続自体は成功し、who_am_i や search_files_keyword、copy_hub など30個のツールは確認・利用できているのですが、ドキュメント等に記載のある ai_qa_hub などのBox AIを活用したQ&Aツールが表示（利用）できません。つきましては、これらAI系のMCPツールを有効化するために必要な条件について教えていただけますでしょうか。具体的には以下の点について確認させてください。 ライセンスの要件： 現在のプラン以外に上位ライセンス（Enterprise Plus等）や、特定の追加オプションが必要でしょうか。 管理コンソールの設定： 管理者側で「Box AI」の有効化など、事前の管理設定が必要でしょうか。 開発者コンソールのスコープ設定： カスタムアプリ（OAuth 2.0）側で、特定のスコープ（例: AI関連の権限）を追加で付与する必要はありますでしょうか。 現状、当環境で確認できているツールのリストは以下の通りです。 （※ここに提示した1〜30のリストを必要に応じて添付、または「基本操作やHub操作を含む30個」と記載）お手数をおかけしますが、ご確認のほどよろしくお願いいたします。

I am trying to use the Box MCP Server Remote with a tool that connects to Google Gemini / Vertex AI API. Authentication seems to be fine however, the connection is failing with a FIELD_INVALID error. Gemini's API requires all MCP tool/function names to be under 64 characters and contain no special characters. It appears one of the tools being provided by the Box Remote MCP server violates this naming convention (specifically the tool at index 28 in the payload). Could the remote server's tool names be updated to comply with the <64 character limit so it can be used with Gemini?Specific Gemini response:Something went wrong. Here's the specific error message we encountered: An error occurred while processing the request: [{ "error": { "code": 400, "message": "The GenerateContentRequest proto is invalid:\n * tools[0].function_declarations[28].name: [FIELD_INVALID] Invalid function name. Must start with a letter or an underscore. Must be a-z, A-Z, 0-9, or contain underscores (_), dots (

Claude is able to edit folders in BOX, however it is not able to edit file names due to “external collaborators.” Has anyone else run into this issue or know how to resolve?

We are using the CLI to test a process of extracting meta data from documents. The Meta data template pulls the data correctly via the Box sidebar. When we run the script via the CLI we get different results. I assume this is due to the Model the CLI is using vs the model the Box site is using. Is there a way to change the CLI to use what the site is using?

Hey Box AI Dev Community,I'm an independent dev working on Box AI integrations & looking to test some of advanced AI features for potential enterprise implementations. Currently hitting some access limitations blocking my dev work.What I'm trying to test:MCP (Model Context Protocol) integration w/ Box Enhanced Extract Agent functionality Box AI Studio configurations General AI agent dev workflowsAccess barriers I'm encountering:Need AI Model permissions enabled Enhanced Extract endpoint access Enable Box AI in Admin Console setting 2FA requirements for certain AI features Box AI Studio access (if available for dev)Questions for the community:Has anyone successfully gotten developer/sandbox access to these enterprise AI features? Is there a specific process for requesting elevated permissions for testing purposes? Any workarounds or alternative approaches for evaluating Box's AI capabilities?I understand these are typically enterprise-level features, but hoping there might be develo

When I upload a .msg (Outlook) to my box, and click to preview, the preview shows only unformatted text sans any images or typography. Is there a way for the preview to show a more representative view of the email, complete with fonts and images?