MCP App Store
by KeenEthics
Egnyte icon

Egnyte MCP Server for Claude and ChatGPT

By Egnyte

Overview

Work with documents and files stored in Egnyte directly from ChatGPT. Search across folders, retrieve relevant files, and extract key information to answer questions or generate summaries. Quickly surface insights from contracts, reports, and internal documents. Combine information from multiple files to create concise summaries, stakeholder updates, or draft written content grounded in your team’s internal materials.

How to connect

  1. Claude: Open the Egnyte listing in the Claude connectors directory and choose Connect. Claude listing Egnyte exposes 25 tools here.
  2. ChatGPT: Open the Egnyte listing in the ChatGPT app directory and choose Connect. ChatGPT listing Egnyte exposes 25 tools here.

Tools

ask_ai_assistant

ChatGPT
Ask a question to Egnyte AI Assistant with optional context from specific files/folders. REQUIRED: Provide 'intent' parameter (max 15 words) explaining WHY you're calling this tool.

ask_document

ChatGPT
Ask a question about a specific document in Egnyte using Egnyte AI. REQUIRED: Provide 'intent' parameter (max 15 words) explaining WHY you're calling this tool.

ask_knowledge_base

ChatGPT
Ask a question to a specific Knowledge Base. REQUIRED: Provide 'intent' parameter (max 15 words) explaining WHY you're calling this tool.

create_comment

ChatGPT
Create a new comment (note) on an existing file in Egnyte. REQUIRED: Provide 'intent' parameter (max 15 words) explaining WHY you're calling this tool.

create_folder

ChatGPT
Create a new folder in Egnyte. Fails if folder already exists or parent folders do not exist. Use list_filesystem_by_path first to verify the parent path exists. REQUIRED: Provide 'intent' parameter (max 15 words) explaining WHY you're calling this tool.

fetch

ChatGPT
Fetch content of a specific document in Egnyte. REQUIRED: Provide 'intent' parameter (max 15 words) explaining WHY you're calling this tool.

get_agent_ask_status

ChatGPT
Check the status of an Agent ask request. Call repeatedly with the agent_id and request_id while status is PLANNING or IN_PROGRESS — response generation can take up to 2 minutes. When status is COMPLETED, read responseText and citations. When status is AWAITING_USER_CONFIRMATION, stop polling — this is a terminal state: the agent is waiting on a confirmation that cannot be given through MCP tools, and responseText is typically empty or omitted from the response. When status is FAILED or COMPLETED_FAILED, stop polling and report the failure to the user. Avoid polling too frequently as excessive requests may result in rate limiting. REQUIRED: Provide 'intent' parameter (max 15 words) explaining WHY you're calling this tool.

get_comment

ChatGPT
Get detailed information for a specific comment in Egnyte by its unique identifier. REQUIRED: Provide 'intent' parameter (max 15 words) explaining WHY you're calling this tool.

get_download_url

ChatGPT
Generate a single-use signed HTTPS URL for direct binary file download. Tokens expire approximately 5 minutes after minting. Do not batch-mint multiple tokens unless you can use each one within that window. RESPONSE FIELDS: - url (string): signed HTTPS URL to fetch the file from. - token (string): single-use authorization token; include as "Authorization: OneTimeDownload <token>". - validUntil (string): ISO-8601 datetime string (e.g. "2026-08-14T12:34:56Z") marking the 5-minute expiry upper bound. - singleUse (boolean): always true; the token is consumed on the first GET and cannot be reused. - fileSize (integer, nullable): file size in bytes; omitted for files where size is unavailable. - fileExtension (string, nullable): file type (e.g. "pdf", "docx"); use to decide binary vs text handling. - checksum (string, nullable): SHA-512 hash of the complete file content; use for integrity verification. - checksumAlgorithm (string, nullable): always "sha512" when checksum is present. PREFER get_file_content FIRST: before calling this tool, check whether get_file_content can satisfy the request — it is simpler, requires no HTTP client, and handles most text-based files directly. Use get_download_url only when get_file_content cannot be used or when the raw binary stream is explicitly required. AGENT ACTION REQUIRED: After calling this tool, YOU (the agent) MUST immediately fetch the file yourself using any available HTTP client (Bash curl, a fetch MCP tool, or any other HTTP tool). Do NOT pass the URL or token to the user — fetch the file and return its content directly. Make a GET request to the returned url with the header: Authorization: OneTimeDownload <token> BINARY RESPONSE — the HTTP response body is raw binary data, not text. You MUST handle it appropriately based on the fileExtension field: - Text formats (txt, csv, json, xml, md, log…): read as UTF-8 and return the content directly. - Binary formats (pdf, docx, xlsx, pptx, jpg, png, mp4, zip…): save the raw bytes to a temporary file and work from there (e.g. extract text, display, or pass the path to the user). Do NOT attempt to decode binary bytes as text — it will produce garbage. CRITICAL CONSTRAINTS — read before use: - Single use: the token is consumed on first GET — singleUse: true in the response confirms this. Never retry the same URL; call get_download_url again to get a fresh token. - validUntil is an ISO-8601 datetime string (e.g. "2026-08-14T12:34:56Z") representing a 5-minute upper bound; a consumed token is dead before that time regardless of validUntil. - Large files over unreliable connections may fail mid-download and require a full re-mint. RANGE REQUESTS — supported; server returns HTTP 206 Partial Content: - Each Range request consumes the token — call get_download_url again before each subsequent Range request. - The download URL is stable per file across get_download_url calls (derived from the file group_id) — only the token changes. For chunked downloads, reuse the URL and swap only the token. - Multi-range downloads are NOT atomic: if the file is modified between token mints, chunk boundaries will be wrong and the assembled content corrupt. Compare the x-sha512-checksum response header across chunks; if it changes, restart from byte 0 with a fresh token. CHECKSUM — for download integrity verification: - checksum is the hash of the complete file (not of an individual Range chunk). Compare it against the full assembled content after all bytes are received. - checksumAlgorithm identifies the hash algorithm (e.g. sha512). - checksum and checksumAlgorithm are included when available. - For chunked Range downloads, also compare the x-sha512-checksum response header across chunks (see RANGE REQUESTS above). ERRORS: - PATH_IS_A_FOLDER: path points to a folder, not a file. Use list_filesystem_by_path to browse the folder, or pass the path to a specific file within it. REQUIRED: Provide 'intent' parameter (max 15 words) explaining WHY you're calling this tool.

get_file_content

ChatGPT
Get text content of a file from Egnyte. Returns markdown/plain-text representation with pagination support. Provide either 'group_id' (group_id) or 'path', not both. For files without extractable text (e.g. PDFs, PPTX, XLSX), use ask_document instead. REQUIRED: Provide 'intent' parameter (max 15 words) explaining WHY you're calling this tool.

get_upload_url

ChatGPT
Generate a short-lived, single-use signed HTTPS URL for direct binary file upload. Returns a structured result with three fields: url (the upload endpoint), token (single-use authorization credential), and validUntil (ISO-8601 upper bound). UPLOAD INSTRUCTION: Use the returned url and token to upload the file via any HTTP client that supports multipart form upload. Example with curl: curl -H "Authorization: OneTimeUpload <token>" -F "file=@localfile" "<url>" The multipart field name MUST be "file". If the target file extension is not known, ask the user before calling this tool — the extension cannot be changed after upload without creating a new version. CRITICAL CONSTRAINTS — read before use: - Single-shot, fail-closed: the token is consumed on the first authorization attempt, even if the upload fails or is interrupted. A second attempt to the same url will fail. - No auto-retry: never retry the same url. If the upload fails, call get_upload_url again to mint a fresh token. - Advisory cap on large uploads: no size limit is enforced at URL generation time, but very large uploads may fail mid-transfer due to a load balancer timeout on the upload endpoint — the connection is dropped without an HTTP response and the file is not committed. For reliable results, keep uploads well under 200 MB. - Existing path = new version: uploading to a path that already exists creates a new file version (non-destructive). - validUntil is an ISO-8601 upper bound — the token may be dead before that time if it was already consumed. Do not use validUntil to decide whether to retry — a consumed token is dead regardless of the timestamp. On any failure, call get_upload_url again for a fresh token. Errors are surfaced as MCP tool failures: invalid input and upstream failures are returned through the normal MCP error path, so handle the call as failed rather than looking for an error field in the result payload. REQUIRED: Provide 'intent' parameter (max 15 words) explaining WHY you're calling this tool.

list_comments

ChatGPT
List comments for a specific file or all comments in the Egnyte domain. REQUIRED: Provide 'intent' parameter (max 15 words) explaining WHY you're calling this tool.

list_filesystem_by_path

ChatGPT
List files and folders in Egnyte file system by path with comprehensive metadata and options. Optionally include folder statistics. REQUIRED: Provide 'intent' parameter (max 15 words) explaining WHY you're calling this tool.

list_knowledge_bases

ChatGPT
List available Knowledge Bases in Egnyte. REQUIRED: Provide 'intent' parameter (max 15 words) explaining WHY you're calling this tool.

list_metadata_namespaces

ChatGPT
Returns all available metadata namespaces and their field definitions for the current domain. Always call this before using set_file_metadata to ensure you use valid namespace keys and field names.Hint: Namespaces are stable; responses may be cached for the duration of the MCP session. REQUIRED: Provide 'intent' parameter (max 15 words) explaining WHY you're calling this tool.

list_projects

ChatGPT
List all project folders in the Egnyte domain. Optionally filter by folderId path and include detailed project information. REQUIRED: Provide 'intent' parameter (max 15 words) explaining WHY you're calling this tool.

set_file_metadata

ChatGPT
Set custom metadata on a file in Egnyte. IMPORTANT: This REPLACES all existing metadata in the specified namespace - it does not merge or append. To preserve existing metadata, first retrieve current values using list_filesystem_by_path with list_custom_metadata=true, then include them in the values parameter. ask user what is desired result. Provide either group_id (UUID) or entry_id (UUID) to identify the file (entry_id takes precedence if both provided). REQUIRED: Provide 'intent' parameter (max 15 words) explaining WHY you're calling this tool.

summarize_document

ChatGPT
Generate a summary of a specific document in Egnyte using Egnyte AI. REQUIRED: Provide 'intent' parameter (max 15 words) explaining WHY you're calling this tool.

upload_file

ChatGPT
Upload a text file to Egnyte. Creates a new file or adds a new version if file already exists at the path. Fails if parent folders do not exist. If the user does not specify a complete path, ask them to clarify the target folder before uploading. Maximum file size is 8MB. Supports plain text content only (UTF-8). Suitable file types: txt, csv, json, xml, rtf, md, html, css, js, java, py, etc. REQUIRED: Provide 'intent' parameter (max 15 words) explaining WHY you're calling this tool.

whoami

ChatGPT
Return the Egnyte identity this connector is authenticated as: the Egnyte username, domain (workgroup), and email address. USE THIS TOOL WHEN: - You need to identify which Egnyte account is connected or verify it is correct. Example: "Which account are you using?", "Am I connected to the right account?" - You need to know what domain or workgroup this connector is bound to. Example: "What domain is this connector bound to?" - You need the email address of the authenticated user. Example: "What is your email address?" USE A DIFFERENT TOOL WHEN: - No competing tools in this server. RESPONSE: - username — the Egnyte login name - domain — the workgroup (tenant) this connector is bound to - email — the account's email address CONSTRAINTS/LIMITATIONS/CAVEATS: - No caveats worth noting. REQUIRED: Provide 'intent' parameter (max 15 words) explaining WHY you're calling this tool.

ask_agent

Claude

ask_ai_assistant

Claude

ask_copilot

Claude

ask_document

Claude

ask_knowledge_base

Claude

create_comment

Claude

create_folder

Claude

download_file_by_id

Claude

download_file_by_path

Claude

fetch

Claude

get_agent_ask_status

Claude

get_comment

Claude

get_download_url

Claude

get_file_content

Claude

get_folder_stats

Claude

get_project_details

Claude

get_upload_url

Claude

get_workflow_details

Claude

list_agents

Claude

list_comments

Claude

list_filesystem_by_id

Claude

list_filesystem_by_path

Claude

list_knowledge_bases

Claude

list_metadata_namespaces

Claude

list_projects

Claude

list_workflow_tasks

Claude

list_workflows

Claude

search_projects_by_folder

Claude

set_file_metadata

Claude

summarize_document

Claude

upload_file

Claude

whoami

Claude

Capabilities

Writes

Example Prompts

Click any prompt to copy it.

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/egnyte"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/egnyte)