mcp

Let your AI app run your socials. You approve.

MCP is the open standard AI apps use to call outside tools. Add one URL to Claude, ChatGPT, Cursor or any MCP client, and it can write, schedule and answer comments in your PostEveryPlatform workspace.

server URL https://posteveryplatform.com/mcp

Works on every plan, Free included. Your plan's limits still apply.

1Your AI app writesDrafts, schedules and replies through the tools below.
2You approveScheduled posts wait in PostEveryPlatform until you say yes.
3It goes out on timeTo the accounts and at the time you approved.

Connect your AI app

Every app below finds our sign-in page by itself. You log in to PostEveryPlatform, pick a workspace and tick what the app may do. It never sees your password.

Claude claude.ai and Claude Desktop

Open Customize → Connectors, click + Add, then Add custom connector.
Name it PostEveryPlatform, paste the server URL and click Continue.
Under Authentication pick Sign in now. Keep the recommended OAuth client option, then click Add.
Log in to PostEveryPlatform in the window that opens and choose what Claude may do.
In a chat, open + → Connectors and switch PostEveryPlatform on.

On Team and Enterprise plans an owner adds it once under Organization settings → Connectors, then each member clicks Connect. Claude's Free plan allows one custom connector.

Claude Code

Add the server. Put --scope user after add to use it in every project.

terminal

claude mcp add --transport http posteveryplatform https://posteveryplatform.com/mcp

Sign in: type /mcp inside Claude Code, or run claude mcp login posteveryplatform. Your browser opens our consent page.

ChatGPT

Go to chatgpt.com/plugins, click the plus button, then Add custom MCP server.
Give it a name and a short description. Under Connection, paste the server URL.
Set authentication to OAuth, read the risk warning and tick I understand and want to continue.
Select Create as a plugin, check the tools it found, and sign in to PostEveryPlatform when asked.

Account and workspace policies apply: in some ChatGPT workspaces an admin has to allow custom MCP servers first.

Cursor

Add this to ~/.cursor/mcp.json (every project) or .cursor/mcp.json (one project):

json

{
  "mcpServers": {
    "posteveryplatform": { "url": "https://posteveryplatform.com/mcp" }
  }
}

Cursor signs in with OAuth when you first use it. Switch the server on or off from Customize in the sidebar.

Any MCP client

Add a remote MCP server (Streamable HTTP) with the URL https://posteveryplatform.com/mcp. Clients that support MCP sign-in discover it from that URL and open our consent page. If yours cannot sign in, use an API key instead.

Our consent page shows the app's name and the address it returns to, then asks three things.

Which workspace. An app works in one workspace. Connect it again to use another.
What it may do. Read, drafts, scheduling and replies are ticked for you. Posting right away is not.
Whether its posts skip your approval. Off unless you tick it.

Permission	What it allows	Ticked at first
`read`	See accounts, posts, ideas, stats and comments	Yes
`draft`	Write drafts, ideas and media, and edit or delete posts that have not gone out. Nothing is sent.	Yes
`schedule`	Schedule posts and use your posting queue. They wait for your approval unless you allow otherwise.	Yes
`publish`	Post right away. Still held for your approval while approval is on.	No
`reply`	Reply to, like or hide comments. Replies go out without approval.	Yes

Scheduling and posting include writing drafts, and every permission includes reading the workspace, so the app can work at all.

No permission reaches billing, your login, connecting or disconnecting social accounts, or approving posts. Those stay with you in the app.

The approval rule

Nothing goes out until you approve it.

When an AI app schedules a post or asks to post now, the post waits in PostEveryPlatform with the status pending_approval. You approve or reject it in the app. Approved posts leave at their time, or within a minute if that time has passed.

draftSaved. Nothing is sent.

pending_approvalWaiting for you.

scheduledApproved. Leaves on time.

Approval is on by default for every workspace. You can switch it off for the whole workspace in Settings → AI agents, or tick Posts skip my approval for one app you trust. Replies to comments are not held: give an app the reply permission only if you are happy for it to answer people.

Tools

What a connected app can call. It only sees the tools its permissions allow. Times are in your workspace's timezone.

Tool	What it does	Needs
Workspace and posts
`get_workspace`	Workspace, timezone and current time, plan limits, posts this month, the approval rule, the app's permissions	read
`list_accounts`	Connected social accounts and their status	read
`list_posts`	Posts in a time window, every draft and every post waiting for approval	read
`get_post`	One post with its accounts, links, errors and your note if you sent it back	read
`create_post`	Write a draft, schedule it, put it in your queue or post now (held for approval)	draft
`update_post`	Edit a post that has not gone out	draft
`delete_post`	Delete a draft, a held post or a scheduled post. Never one that went out.	draft
`retry_post`	Send a failed post again to the accounts that failed (only where its posts skip your approval)	publish
`upload_media`	Add an image or video from a public https link, up to 95 MB	draft
Ideas and planning
`list_ideas`	Your ideas board, with search	read
`create_idea`	Add an idea to the board	draft
`update_idea`	Edit an idea or move it to drafting or ready	draft
`list_feed_items`	Your RSS feeds and the newest items they brought in	read
`get_next_slots`	The next free times in your posting schedule	read
`get_schedule`	Your weekly posting times and whether the queue is paused	read
Results
`get_analytics`	Totals for 7 to 90 days: published, failed, views, engagement, streak, top posts	read
`get_post_analytics`	Views, likes, comments and shares per post and platform	read
`get_best_times`	The weekday and hour your posts do best, once 10 posts have numbers	read
Comments
`list_comments`	Comments on your recent posts, newest first	read
`reply_to_comment`	Reply as the account the comment was left on. Replies are not held for approval.	reply
`comment_action`	Like or hide a comment where the platform allows it	reply

Use an API key instead

For scripts, servers and clients without MCP sign-in, make a key in Settings → AI agents and send it as a header. A key carries the permissions and approval setting you gave it, like an app.

header

Authorization: Bearer pep_…

With Claude Code:

terminal

claude mcp add --transport http posteveryplatform https://posteveryplatform.com/mcp \
  --header "Authorization: Bearer pep_…"

Treat a key like a password: anyone holding it can do what it allows. Delete it in the app when you no longer need it.

Limits

Requests: 30 a minute per app on Free, 120 a minute on paid plans. Over that, the app is told how many seconds to wait.
Posts and accounts: Your plan's limits, the same as in the app. Free sends 15 posts a month to 3 accounts.
Sign-in: Access lasts 1 hour and renews by itself. An app unused for 30 days signs in again.
Media: Images (JPG, PNG, WebP, GIF) or one video (MP4, MOV, WebM), up to 95 MB per file from a link.

Disconnect an app

Open AI agents in the app. Connected apps are listed with their permissions and when they were last used. Disconnect ends the app's access at once, on every device it signed in from.

Removing the server inside your AI app (for example claude mcp remove posteveryplatform) deletes that app's copy of the sign-in. To be sure it can never connect again, disconnect it in PostEveryPlatform too.

For developers

The server follows the MCP specification revision 2026-07-28 and still answers clients that open with initialize (2025-11-25, 2025-06-18, 2025-03-26). There are no sessions. Responses are plain JSON; subscriptions/listen is acknowledged and closed, because the tool list does not change while you are connected.

Protected resource metadata: /.well-known/oauth-protected-resource
Authorization server metadata: /.well-known/oauth-authorization-server
Client identity: Client ID Metadata Documents, or dynamic registration at /oauth/register. Public clients only.
Sign-in: Authorization code with PKCE (S256), the resource parameter, and iss in every response.
Tokens: Refresh tokens rotate on every use. Using an old one ends the connection. Revoke at /oauth/revoke.
Redirect URIs: https, or http on localhost / 127.0.0.1 / [::1] on any port.