---
title: "Connect FeedMCP to DeepSeek (Social Media Data MCP & API) | FeedMCP"
description: "Pair DeepSeek-V3 or DeepSeek API with FeedMCP via **Cherry Studio, Cline, Cursor or OpenAI-compatible Function Calling** to read all 21 social platforms at"
canonical: "https://feedmcp.com/integrations/deepseek/"
---
<p class="crumbs"><a href="/integrations/">Integrations</a> / DeepSeek</p>

# Connect FeedMCP to DeepSeek (Social Media Data MCP & API)

Pair DeepSeek-V3 or DeepSeek API with FeedMCP via **Cherry Studio, Cline, Cursor or OpenAI-compatible Function Calling** to read all 21 social platforms at pay-per-call prices.

| Item | Value |
| --- | --- |
| Method | MCP client (Cherry Studio / Cline) or Function Calling |
| MCP URL | `https://api.feedmcp.com/mcp` |
| OpenAPI | https://feedmcp.com/openapi.json |
| Coverage | 21 platforms · 193 endpoints |
| Billing | Pay per call (from $0.0003), failures free, $0.10 trial credit |

## Setup steps

### 1. Create a spend-capped API key

Sign up in the [console](/console/keys/) ($0.10 free trial credit, no card) and create a dedicated API key (`fmk_live_…`) for DeepSeek with a spend cap.

### 2. Configure FeedMCP in DeepSeek

**Option 1 (no-code, recommended)**: Configure your DeepSeek API key inside [Cherry Studio](/integrations/cherry-studio/) and enable the FeedMCP remote MCP server (`https://api.feedmcp.com/mcp`).

**Option 2 (code / Function Calling)**: Wire a thin FeedMCP HTTP wrapper into DeepSeek's `tools` array:

```python
import os, requests

def call_feedmcp(endpoint_id: str, params: dict, max_price: str = "0.02"):
    path = "/v1/" + endpoint_id.replace(".", "/")
    r = requests.get(
        f"https://api.feedmcp.com{path}",
        params=params,
        headers={
            "Authorization": f"Bearer {os.environ['FEEDMCP_API_KEY']}",
            "X-Max-Price": max_price,
        },
        timeout=30,
    )
    return r.json()
```

### 3. Verify the connection

Test `GET https://api.feedmcp.com/v1/catalog/search?q=tiktok` (free, no key) first, then run a $0.002 trending call with your key.

## First prompts to try

| Prompt | Estimated cost |
| --- | --- |
| Fetch today's top 10 trends on Douyin, Weibo and Zhihu and write a daily briefing | 3 calls, $0.006 |
| Search Xiaohongshu for home espresso machine notes and group buyer pain points | 1 call, $0.02 |
| Pull the last 10 posts from this WeChat Official Account and analyze winning titles | ~$0.22 |

## Tips & guardrails

Combining DeepSeek with FeedMCP keeps both LLM token cost and per-call social data cost extremely low for batch Chinese and global analysis.

## Questions

### How much does FeedMCP cost in DeepSeek?

Pay per call with no subscription, starting at $0.0003 (most platforms $0.002/call; Xiaohongshu and WeChat $0.02/call). Failed calls are free; $0.10 free trial credit on sign-up.

### How do I prevent DeepSeek from overspending?

Set a total spend cap on the API key in the console, and use `X-Max-Price` / `max_price` to cap individual calls.

### Which platforms can DeepSeek read through FeedMCP?

All 21 platforms and 193 endpoints: TikTok, YouTube, Instagram, X, Threads, Telegram, Reddit, LinkedIn, Facebook, Pinterest, Meta Ad Library, Google Ads Transparency, Xiaohongshu (RedNote), Douyin, Kuaishou, WeChat Official Accounts, WeChat Channels, WeChat Search, Weibo, Zhihu and Bilibili. See [all endpoints](/apis/).

## Other AI client guides

- [Claude Code](/integrations/claude-code/)：Native remote MCP (Streamable HTTP)
- [Claude (Desktop / Web)](/integrations/claude/)：Remote MCP server (Streamable HTTP)
- [Cursor](/integrations/cursor/)：Native remote MCP (`~/.cursor/mcp.json`)
- [OpenAI Codex CLI](/integrations/codex/)：Native remote MCP (`~/.codex/config.toml`)
- [Windsurf](/integrations/windsurf/)：Native remote MCP (`mcp_config.json`)
- [Cherry Studio](/integrations/cherry-studio/)：GUI remote MCP (Streamable HTTP)
- [ChatGPT (connectors / Custom GPT Actions)](/integrations/chatgpt/)：Remote MCP connector (OAuth sign-in), or Custom GPT Actions (`/openapi.json`)
- [WorkBuddy](/integrations/workbuddy/)：Remote MCP (Streamable HTTP) or `agents.md` setup
- [Doubao & Coze](/integrations/doubao/)：Coze plugin / OpenAPI import + MCP
- [Qwen (Tongyi / Model Studio)](/integrations/qwen/)：Remote MCP (Streamable HTTP) or HTTP tool calling
- [Dify & n8n Workflows](/integrations/dify-n8n/)：OpenAPI import / HTTP Request node / MCP node

---

HTML version: https://feedmcp.com/integrations/deepseek/
