Quickstart: DeepSeek

DeepSeek is a model, not an MCP host. The DeepSeek app/website has no custom-connector slot, so you can’t add Pipeworx to “DeepSeek” directly the way you can with Claude, ChatGPT, or Grok. That’s not a Pipeworx limitation — DeepSeek simply ships a model and an API, not a connector surface.

The way to use DeepSeek with Pipeworx is to run a DeepSeek model inside an MCP-capable client and point that client at Pipeworx.

How to use it

  1. Pick a client that lets you choose your model and add MCP servers — for example:
  2. Set the model/provider to DeepSeek. DeepSeek’s API is OpenAI- and Anthropic-compatible, so any of these clients can drive it — drop in your DeepSeek API key and select a DeepSeek model.
  3. Add Pipeworx in that client using its quickstart, with the URL:
https://pipeworx.io/mcp

Now DeepSeek (the model) calls Pipeworx (the tools) through the client. Follow the client’s quickstart for the exact connect + teach steps.

Why no direct connector?

MCP connectors are hosted by the client (the app/CLI/IDE), not the model. Pipeworx connects to clients; the client then runs whatever model you choose — DeepSeek included. Tiers and limits are the same as everywhere else: anonymous 50/day · free account 200/day · Paid unlimited.

Last reviewed June 9, 2026