FeedNest MCP Server

Your sources. Not the whole web. Ask your AI about your feeds - it already knows.

ChatGPT searches the internet. Perplexity crawls billions of pages. When you ask about your industry, your interests, your world - they give you everything except what matters.

FeedNest is different. Your AI works exclusively with the sources you choose: the blogs you trust, the publications you follow, the experts you believe in. No algorithmic noise. No black-box results. Intelligence grounded in your world.

This MCP server gives any AI assistant - ChatGPT, Codex, Claude, Gemini, Grok, Cursor, Windsurf, Cline, Le Chat, and more - direct access to your feeds, articles, highlights, notes, and tags through natural language.

What Your AI Can Do

Just ask. Your AI handles the rest.

• Morning briefings - Wake up to a summary of everything that happened overnight across your feeds

• Compare coverage - See how different sources report the same story, side by side

• Spot trends early - Your AI detects emerging patterns before they hit mainstream

• Deep-dive any topic - Get a comprehensive synthesis from every source you follow

• Track developing stories - Follow a narrative as it evolves across all your feeds

• Instant search - Find any article across all your sources in seconds

• Highlights and notes - Save what matters and annotate with your own thoughts

Related MCP server: Inoreader MCP Server

Quick Start

Prerequisites

1. A FeedNest account with a Pro subscription

2. An API key - generate one from Settings → Developer API in your dashboard

ChatGPT

Option A: Web UI (OAuth - recommended)

1. Go to Settings → Apps & Connectors → Advanced and enable Developer Mode

2. Go to Settings → Apps → Create. Name it FeedNest and set the URL to https://mcp.feednest.com

3. ChatGPT will redirect you to FeedNest - log in and authorize

No API key needed. ChatGPT handles the OAuth flow automatically.

Option B: Responses API

Include FeedNest as an MCP tool in your API request:

{
  "type": "mcp",
  "server_url": "https://mcp.feednest.com",
  "server_label": "feednest",
  "authorization": "fn_live_YOUR_API_KEY"
}

Codex

Option A: Plugin (recommended)

The plugin bundles skills, MCP config, and brand assets in one install. It is not in the Codex plugin directory yet, so install it from this repository: see Local Installation in the Codex plugin README.

Option B: MCP in config.toml

Add to ~/.codex/config.toml:

[mcp_servers.feednest]
url = "https://mcp.feednest.com"
bearer_token_env_var = "FEEDNEST_API_KEY"

export FEEDNEST_API_KEY="fn_live_YOUR_API_KEY"

Or via CLI:

codex mcp add feednest --transport http https://mcp.feednest.com

Claude

Connectors directory (OAuth - recommended)

FeedNest is listed in Claude's connectors directory.

1. Open claude.ai/directory/feednest, or search for FeedNest in Customize → Connectors

2. Click Connect to Claude, log in to FeedNest, and authorize

No API key needed, and the connector follows your Claude account to Claude Desktop and the mobile apps.

To add it by URL instead, add a custom connector in Customize → Connectors with the URL https://mcp.feednest.com and leave OAuth Client ID and Secret empty.

Claude Code

claude mcp add --transport http feednest https://mcp.feednest.com \
  --header "Authorization: Bearer fn_live_YOUR_API_KEY"

Add -s user for global access across all projects.

Claude Desktop config file (requires bridge)

The directory connector above already works in Claude Desktop. To configure FeedNest in claude_desktop_config.json instead, which only runs stdio servers, use the mcp-remote bridge:

{
  "mcpServers": {
    "feednest": {
      "command": "npx",
      "args": [
        "-y", "mcp-remote",
        "https://mcp.feednest.com",
        "--header", "Authorization:${AUTH_HEADER}"
      ],
      "env": {
        "AUTH_HEADER": "Bearer fn_live_YOUR_API_KEY"
      }
    }
  }
}

Config path: ~/Library/Application Support/Claude/claude_desktop_config.json (macOS), %APPDATA%\Claude\claude_desktop_config.json (Windows), or ~/.config/Claude/claude_desktop_config.json (Linux).

Cursor

Add to .cursor/mcp.json in your project (or ~/.cursor/mcp.json for global):

{
  "mcpServers": {
    "feednest": {
      "url": "https://mcp.feednest.com",
      "headers": {
        "Authorization": "Bearer fn_live_YOUR_API_KEY"
      }
    }
  }
}

Tip: Cursor supports env var interpolation: "Bearer ${env:FEEDNEST_API_KEY}". Do not add a "type" field - Cursor auto-detects from the "url" key.

Le Chat (Mistral)

Available on all plans, including the free tier.

1. Go to Sidebar → Intelligence → Connectors

2. Click + Add Connector → Custom MCP Connector

3. Set Name to FeedNest, Server URL to https://mcp.feednest.com

4. Set Auth type to API Token Authentication, Authorization to Bearer fn_live_YOUR_API_KEY

Windsurf

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "feednest": {
      "serverUrl": "https://mcp.feednest.com",
      "headers": {
        "Authorization": "Bearer fn_live_YOUR_API_KEY"
      }
    }
  }
}

Important: Windsurf uses "serverUrl" (not "url"). Using the wrong key causes a silent failure.

Cline

In VS Code, open Cline extension settings → MCP Servers configuration (JSON):

{
  "mcpServers": {
    "feednest": {
      "type": "streamableHttp",
      "url": "https://mcp.feednest.com",
      "headers": {
        "Authorization": "Bearer fn_live_YOUR_API_KEY"
      }
    }
  }
}

Critical: The "type": "streamableHttp" field is mandatory for Cline. Without it, custom headers may fail silently and authentication won't work.

Gemini CLI

Add to ~/.gemini/settings.json:

{
  "mcpServers": {
    "feednest": {
      "httpUrl": "https://mcp.feednest.com",
      "headers": {
        "Authorization": "Bearer fn_live_YOUR_API_KEY"
      }
    }
  }
}

Note: Gemini CLI uses "httpUrl" (not "url"). The "url" field is reserved for SSE transport. CLI only - gemini.google.com does not support external MCP servers.

Grok (xAI)

Available through the xAI API/SDK only, not from the grok.com chat UI.

curl https://api.x.ai/v1/responses \
  -H "Authorization: Bearer $XAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "grok-3",
    "tools": [{
      "type": "mcp",
      "server_url": "https://mcp.feednest.com",
      "server_label": "feednest",
      "authorization": "Bearer fn_live_YOUR_API_KEY"
    }],
    "input": "What are my unread articles?"
  }'

Any other MCP client

Any client supporting Streamable HTTP can connect using:

• URL: https://mcp.feednest.com

• Header: Authorization: Bearer fn_live_YOUR_API_KEY

For stdio-only clients, use the mcp-remote bridge:

{
  "mcpServers": {
    "feednest": {
      "command": "npx",
      "args": [
        "-y", "mcp-remote",
        "https://mcp.feednest.com",
        "--header", "Authorization:Bearer fn_live_YOUR_API_KEY"
      ]
    }
  }
}

Tools (25)

Articles

Read Status

Saving

Feeds & Structure

Tagging

Highlights

Notes

History

Output Feeds

AI

text_to_speech is not exposed by the hosted MCP server: it generates audio with an AI model, which the Claude connectors directory does not allow. Reading articles aloud stays available in the FeedNest app and through the OpenClaw plugin.

Stats

* = required parameter

Example Conversation

You: "What's happening in AI today?"

ChatGPT: "Based on your tech feeds, here are today's 3 biggest stories:

1. OpenAI announced GPT-5, covered first in your Hacker News feed
2. EU AI Act enforcement begins, mentioned across 4 of your sources
3. Google published new Gemini research, from your AI folder

Want me to summarize any of them?"

More things you can ask

• "Give me a morning briefing from my feeds"

• "Compare how Bloomberg and Reuters covered this story"

• "What did I read yesterday?"

• "Save this article and highlight the key findings"

• "Tag all AI-related articles with 'Machine Learning'"

• "How's my reading streak going?"

Authentication

API Key (personal use)

Generate a key from Settings → Developer API in the FeedNest dashboard. Keys start with fn_live_ followed by 32 hex characters.

Authorization: Bearer fn_live_abc123...

OAuth 2.1 (third-party apps)

The server supports full OAuth 2.1 with PKCE. Discovery endpoint:

https://mcp.feednest.com/.well-known/oauth-protected-resource

ChatGPT and Claude.ai handle OAuth automatically - no API key needed for those.

Scopes

You Stay in Control

• You choose what's visible - Pick exactly which feeds, folders, and actions your AI can access

• Encrypted and verified - Every request is secured so no one else can reach your data

• Built-in usage limits - 1,000 requests/hour per key, your account stays protected

• Revoke access instantly - One click from your dashboard

Troubleshooting

Packages

This repository is a monorepo containing the client libraries and plugins for the FeedNest MCP server.

Development

npm install
npm run build    # Build all packages (sdk → shared-tool-defs → openclaw-plugin)
npm test         # Run all tests
npm run type-check

Links

• FeedNest - the app

• Documentation - full docs with setup guides per platform

• Integrations - see all AI integrations

License

MIT