Windsurf with CustomGPT.ai’s Hosted MCP (Model Context Protocol) Server – Connect Your Private Data 

Yes. You can connect Windsurf to a hosted MCP endpoint by generating a token in your project, copying the JSON configuration, and pasting it into Windsurf’s custom server settings. Kevin Petrie described CustomGPT.ai as “a no-code platform for creating custom AI business agents,” which fits the appeal of a hosted MCP setup: you can use private knowledge in Windsurf without running and maintaining your own MCP server. If you prefer, self-hosting is still an alternative, but it is not required for this workflow.

You can reduce risk by limiting each project to the documents you actually want searchable, issuing a dedicated token, and reviewing access before sharing the connection. The documented setup also points to a hosted service that is SOC 2 Type 2 certified, GDPR compliant, and states that uploaded data is not used for model training. For sensitive use cases, keep project scope narrow so Windsurf only retrieves from the content you intended to expose.

Open Windsurf and go to Settings → Advanced → Cascade. Then choose Add Server → Add custom server +, paste the full JSON block generated from Deploy → MCP Server (Beta) in your project, and save it. When the server status light turns green, Windsurf is connected and can query your indexed private data through that MCP endpoint.

The intended flow is retrieval first. The MCP server finds relevant information in your project and sends that context back to Windsurf, so answers are grounded in your files instead of relying only on model memory. This is a RAG workflow, and the platform also supports anti-hallucination with citation support. Its published RAG benchmark reports higher accuracy than OpenAI on that test, but answer quality still depends on how well your documents are organized and indexed.

Yes. Biamp used the platform for both customer support and HR, rolled it out in under 30 days, and supports 90+ languages. Toyon Nurul Huda, Data Scientist at Biamp, said, “CustomGPT has opened new doors for how Biamp interacts with customers and internal audiences. With its advanced GPT-4 capabilities, CustomGPT allows Biamp to quickly address the most common questions and requests for information, making it far faster and more efficient to deliver answers.” In Windsurf, that same pattern can support onboarding, policy lookup, product documentation, and internal research if the right files are indexed first.

The documented setup confirms support for content that has already been ingested into your project, including documents, URLs, JSON, CSV, HTML, XML, audio, and video. The provided materials do not verify a native SharePoint, OneDrive, or database connector for this specific Windsurf workflow. If direct access to one of those systems matters, first confirm whether your MCP-compatible tooling exposes it, or ingest the needed files into your project before using Windsurf.

For many teams, yes. Bill French said, “They’ve officially cracked the sub-second barrier, a breakthrough that fundamentally changes the user experience from merely ‘interactive’ to ‘instantaneous’.” In this workflow, Windsurf connects directly to the hosted MCP endpoint and receives answers back in the app, so you avoid the overhead of maintaining your own server. Actual speed still depends on project size, document quality, and network conditions, but the available evidence points to responsiveness being a core strength.