Integrations / OpenAI Codex CLI
Connect FeedMCP to OpenAI Codex CLI (Social Media Data MCP & API)
Connect FeedMCP to OpenAI Codex CLI with 3 lines in ~/.codex/config.toml so Codex can query 193 social data endpoints right from your terminal.
| Item | Value |
|---|---|
| Method | Native remote MCP (~/.codex/config.toml) |
| 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 ($0.10 free trial credit, no card) and create a dedicated API key (fmk_live_…) for OpenAI Codex CLI with a spend cap.
2. Configure FeedMCP in OpenAI Codex CLI
Add to ~/.codex/config.toml:
[mcp_servers.feedmcp]
url = "https://api.feedmcp.com/mcp"
http_headers = { Authorization = "Bearer fmk_live_YOUR_KEY" }
3. Verify the connection
Start Codex and ask "Use feedmcp to find YouTube transcript endpoints"—it should return youtube.videos.transcript.get at $0.002 for free.
First prompts to try
| Prompt | Estimated cost |
|---|---|
| Pull today's top 10 trends on X and TikTok via FeedMCP and save them as JSON | 2 calls, $0.004 |
| Get the transcript of this YouTube video and write timestamped notes | 1 call, $0.002 |
| Search TikTok Shop for "coffee grinder" and summarize buyer complaints | 2 calls, $0.004 |
Tips & guardrails
Keep the header in your user-level ~/.codex/config.toml rather than inside a shared git repository.
Questions
How much does FeedMCP cost in OpenAI Codex CLI?
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 OpenAI Codex CLI 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 OpenAI Codex CLI 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.
Other AI client guides
- Claude Code:Native remote MCP (Streamable HTTP)
- Claude (Desktop / Web):Remote MCP server (Streamable HTTP)
- Cursor:Native remote MCP (
~/.cursor/mcp.json) - Windsurf:Native remote MCP (
mcp_config.json) - Cherry Studio:GUI remote MCP (Streamable HTTP)
- ChatGPT (Custom GPT Actions):Custom GPT Actions (
/openapi.json) - DeepSeek:MCP client (Cherry Studio / Cline) or Function Calling
- WorkBuddy:Remote MCP (Streamable HTTP) or
agents.mdsetup - Doubao & Coze:Coze plugin / OpenAPI import + MCP
- Qwen (Tongyi / Model Studio):Remote MCP (Streamable HTTP) or HTTP tool calling
- Dify & n8n Workflows:OpenAPI import / HTTP Request node / MCP node