generated: '2026-08-04' method: searched status: published source: https://dataloop.ai/blog/query-dataloop-docs-and-platform-context-using-natural-language-with-the-mcp-server-in-cursor/ announced: '2025-07-02' server: name: Dataloop MCP Server vendor: Dataloop implementation: FastAPI application, deployed as a Dataloop app/pipeline transport: http url: null hosted_endpoint: none-published repository: https://github.com/dataloop-ai-apps/dataloop-mcp install: Add the MCP directly to a Dataloop project on Cursor by copying the pipeline link from the GitHub repository and pasting it into the "Add from GitHub" field. tools: - name: ask-dataloop description: Answer a natural-language question about the Dataloop platform and return a structured response sourced from the official Dataloop documentation (e.g. "How do I create a dataset?", "What are the available pipeline node types?"). input_schema: unknown source: provider announcement clients_documented: - Cursor scope: Documentation question-answering only. The server does NOT expose the Dataloop platform REST surface (projects, datasets, items, annotations, tasks, services) as MCP tools — it is a docs-retrieval tool, so an agent still has to fall back to the SDK or the REST API to act on the platform. x-evidence: - fetched: '2026-08-04' url: https://dataloop.ai/blog/query-dataloop-docs-and-platform-context-using-natural-language-with-the-mcp-server-in-cursor/ http_status: 200 finding: Official Dataloop announcement of the Dataloop MCP Server, naming the single tool `ask-dataloop` and linking the GitHub repository. - fetched: '2026-08-04' url: https://github.com/dataloop-ai-apps/dataloop-mcp http_status: 404 finding: The repository linked from the announcement no longer resolves publicly (private or renamed). Recorded as observed; the server itself is still advertised by the provider. - fetched: '2026-08-04' url: https://mcp.dataloop.ai http_status: 0 finding: TLS handshake failure — no hosted MCP endpoint on that hostname. - fetched: '2026-08-04' url: https://gate.dataloop.ai/mcp http_status: 405 finding: POST tools/list rejected; not an MCP endpoint. probes: tools_list_attempted: true tools_list_result: No publicly reachable HTTP MCP endpoint. tools/list could not be executed against any Dataloop host, so the live inputSchema for `ask-dataloop` is not captured here — only the tool name and description the provider published. gap: summary: Dataloop ships an official MCP server but publishes neither a hosted remote endpoint nor a public repository/registry entry, so the tool manifest cannot be introspected by an agent. Publishing a remote MCP URL (or restoring the public repo) would make the server discoverable rather than announcement-only. deployment: mode: remote verified: searched tools: 1 checked: '2026-08-12' source: catalog MCP census