{"id":1551,"date":"2026-05-27T14:25:20","date_gmt":"2026-05-27T14:25:20","guid":{"rendered":"https:\/\/feeder.co\/help\/?p=1551"},"modified":"2026-09-08T07:29:44","modified_gmt":"2026-09-08T07:29:44","slug":"connect-feeder-to-your-ai-assistant-with-mcp","status":"publish","type":"post","link":"https:\/\/feeder.co\/help\/account\/automation\/connect-feeder-to-your-ai-assistant-with-mcp\/","title":{"rendered":"Connect Feeder to your AI assistant with MCP"},"content":{"rendered":"<p>Feeder&#8217;s MCP server lets Claude, ChatGPT, Codex, Cursor, and other Model Context Protocol clients read your feeds, search posts, and manage collections, straight from your account. Here&#8217;s how to connect it.<\/p>\n<h2>What you&#8217;ll need<\/h2>\n<ul>\n<li>A Feeder Plus, Professional, or Enterprise plan.<\/li>\n<li>An AI client that supports MCP.<\/li>\n<li>The server URL: <code>https:\/\/feeder.co\/mcp<\/code><\/li>\n<\/ul>\n<p>For most clients there are no tokens to copy. However you connect, your client sends you to Feeder to log in and approve access.<\/p>\n<p><strong>One URL to remember.<\/strong> <code>https:\/\/feeder.co\/mcp<\/code> is the only address most clients need. When a setup screen asks for a server URL, an MCP URL, or an endpoint, that is the one to paste. The Authorization URL and Token URL further down this page are different addresses for a different job, and using one of those as the server URL is the most common reason a connection never reaches the approval screen.<\/p>\n<h2>Option 1: Paste the server URL<\/h2>\n<p>Best for clients with a built-in connector screen, like Claude and ChatGPT.<\/p>\n<ul>\n<li>Open your client&#8217;s connectors or integrations settings.<\/li>\n<li>Add a custom connector and paste <code>https:\/\/feeder.co\/mcp<\/code>.<\/li>\n<li>Log in to Feeder when prompted and approve access.<\/li>\n<\/ul>\n<p>In Claude this lives under Settings &gt; Connectors. In ChatGPT it&#8217;s under Settings &gt; Connectors too, and you may need to turn on developer mode first. Custom connectors are a paid feature in both apps.<\/p>\n<h2>Option 2: Paste the config snippet<\/h2>\n<p>Best for Claude Desktop and Cursor. Add this to your client&#8217;s MCP config file:<\/p>\n<pre><code>{\n  \"mcpServers\": {\n    \"feeder\": {\n      \"url\": \"https:\/\/feeder.co\/mcp\"\n    }\n  }\n}<\/code><\/pre>\n<p>Your client will prompt you to log in to Feeder and approve.<\/p>\n<h2>Option 3: Connect from the command line<\/h2>\n<p>Best for Codex CLI and Claude Code, where a single command wires up Feeder.<\/p>\n<h3>Codex CLI<\/h3>\n<pre><code>codex mcp add feeder --url https:\/\/feeder.co\/mcp\ncodex mcp login feeder<\/code><\/pre>\n<p>The login command opens Feeder in your browser, so you can sign in and approve access.<\/p>\n<h3>Claude Code<\/h3>\n<pre><code>claude mcp add --transport http feeder https:\/\/feeder.co\/mcp<\/code><\/pre>\n<p>Then type <code>\/mcp<\/code> inside Claude Code to log in to Feeder and approve.<\/p>\n<h2>Advanced: create a Client ID and secret<\/h2>\n<p>Most clients register themselves when you paste the server URL, so you can skip this section entirely. It is here for the few that can&#8217;t, like Elvex and Azure AI Foundry, which ask for a Client ID and Client Secret up front. For those, generate a pair in Feeder and paste them in.<\/p>\n<ul>\n<li>In Feeder, open your MCP settings under Settings &gt; Account and expand Advanced.<\/li>\n<li>Under Create a Client ID and secret, paste the Redirect URL your client uses, if you have it. Your client&#8217;s setup screen shows it. Add a name if you like.<\/li>\n<li>Select Create Client ID and secret.<\/li>\n<li>Copy the Client Secret right away. For your security, Feeder shows it only once. The Client ID stays available.<\/li>\n<\/ul>\n<p>The pair you create uses <code>client_secret_basic<\/code>, so that is the answer when a client asks how it should send its credentials.<\/p>\n<p>The Redirect URL is optional. If your client only shows it once the Client ID and secret are saved, leave the field empty and add the URL afterwards. A pair without a real Redirect URL can&#8217;t finish an authorization, so it is worth coming back to. Every pair you create is listed under Your Client IDs in the same Advanced section, where you can edit its redirect URL or delete it.<\/p>\n<p>Some clients ask for these two addresses by name. Fill one in only where a field asks for it, and keep using the server URL above everywhere else:<\/p>\n<ul>\n<li>Authorization URL: <code>https:\/\/feeder.co\/1\/oauth2\/authorize<\/code><\/li>\n<li>Token URL: <code>https:\/\/feeder.co\/1\/oauth2\/token<\/code><\/li>\n<li>Scope: <code>mcp<\/code><\/li>\n<\/ul>\n<p>Each Feeder account needs its own Client ID and secret, so create a pair per account.<\/p>\n<h3>Azure AI Foundry<\/h3>\n<p>Foundry only shows you its redirect URL once the connection is saved, so there is nothing to paste into Feeder up front. Do it in this order:<\/p>\n<ul>\n<li>In Feeder, create a Client ID and secret with the Redirect URL field left empty. Copy both values.<\/li>\n<li>In Foundry, add the MCP tool with the server URL <code>https:\/\/feeder.co\/mcp<\/code> and choose OAuth. Paste the Client ID and Client Secret, set the Auth URL and Token URL to the ones above, and set the scope to <code>mcp<\/code>. Foundry marks the Client Secret as optional, but Feeder needs it. If there is a Refresh URL field you can leave it empty, Feeder&#8217;s access tokens last a year.<\/li>\n<li>Save the connection, then copy the Redirect URL Foundry shows. It looks like <code>https:\/\/global.consent.azure-apim.net\/redirect\/<\/code> followed by an ID.<\/li>\n<li>Back in Feeder, find the pair under Your Client IDs, paste that address into its Redirect URL field, and select Save.<\/li>\n<li>Connect in Foundry and approve access when Feeder asks.<\/li>\n<\/ul>\n<p>If Foundry hands you a new redirect URL later, for example after you edit the connection, paste the new one into Feeder the same way.<\/p>\n<h2>If your client never reaches the approval screen<\/h2>\n<p>When the login never opens, or your client loops without ever asking you to approve, it is almost always the server URL.<\/p>\n<ul>\n<li>Check that your client has <code>https:\/\/feeder.co\/mcp<\/code>, and not the Authorization URL or the Token URL.<\/li>\n<li>Remove the Feeder server from your client, add it again, and start the login. In Codex CLI that is <code>codex mcp remove feeder<\/code>, then the two commands in Option 3. In Claude Code, <code>claude mcp remove feeder<\/code>, then add it again. In Claude or ChatGPT, delete the connector and add it back.<\/li>\n<li>If you created a Client ID and secret without a Redirect URL, add the real one under Your Client IDs before you try again.<\/li>\n<\/ul>\n<p>Your client may also tell you Feeder needs a Client Secret. It doesn&#8217;t. Claude, ChatGPT, Codex, Claude Code, and Cursor all register themselves from the server URL alone, and none of them needs anything from the Advanced section.<\/p>\n<p>Still stuck? Write to us at <a href=\"mailto:support@feeder.co\">support@feeder.co<\/a> and tell us which client you&#8217;re using and what the error says.<\/p>\n<h2>Manage connected apps<\/h2>\n<p>Open Feeder and go to Settings &gt; Account to see everything you&#8217;ve connected, and disconnect any app in a click.<\/p>\n<h2>What to ask<\/h2>\n<p>Once you&#8217;re connected, try:<\/p>\n<ul>\n<li>&#8220;What are the headlines from my Tech folder today?&#8221;<\/li>\n<li>&#8220;Find an RSS feed for The Verge and add it to my Mornings folder.&#8221;<\/li>\n<li>&#8220;Save the three most interesting unread posts to my Read Later collection.&#8221;<\/li>\n<\/ul>\n","protected":false},"excerpt":{"rendered":"<p>Connect Claude, ChatGPT, Cursor, or any MCP client to Feeder so your AI assistant can read your feeds, search posts, and manage collections. Requires a Professional or Enterprise plan.<\/p>\n","protected":false},"author":14,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_feature_clip_id":0,"_jetpack_memberships_contains_paid_content":false,"footnotes":"","jetpack_publicize_message":"","jetpack_publicize_feature_enabled":true,"jetpack_social_post_already_shared":true,"jetpack_social_options":{"image_generator_settings":{"template":"highway","default_image_id":0,"font":"","enabled":false},"version":2},"jetpack_post_was_ever_published":false},"categories":[206],"tags":[],"yst_prominent_words":[],"class_list":["post-1551","post","type-post","status-publish","format-standard","hentry","category-automation"],"jetpack_publicize_connections":[],"jetpack_sharing_enabled":true,"jetpack_shortlink":"https:\/\/wp.me\/p893aM-p1","jetpack_featured_media_url":"","_links":{"self":[{"href":"https:\/\/feeder.co\/help\/wp-json\/wp\/v2\/posts\/1551","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/feeder.co\/help\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/feeder.co\/help\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/feeder.co\/help\/wp-json\/wp\/v2\/users\/14"}],"replies":[{"embeddable":true,"href":"https:\/\/feeder.co\/help\/wp-json\/wp\/v2\/comments?post=1551"}],"version-history":[{"count":6,"href":"https:\/\/feeder.co\/help\/wp-json\/wp\/v2\/posts\/1551\/revisions"}],"predecessor-version":[{"id":1625,"href":"https:\/\/feeder.co\/help\/wp-json\/wp\/v2\/posts\/1551\/revisions\/1625"}],"wp:attachment":[{"href":"https:\/\/feeder.co\/help\/wp-json\/wp\/v2\/media?parent=1551"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/feeder.co\/help\/wp-json\/wp\/v2\/categories?post=1551"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/feeder.co\/help\/wp-json\/wp\/v2\/tags?post=1551"},{"taxonomy":"yst_prominent_words","embeddable":true,"href":"https:\/\/feeder.co\/help\/wp-json\/wp\/v2\/yst_prominent_words?post=1551"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}