MCP Server

Connect the MultiPost desktop app to Claude Code / Cursor and other AI tools

Once the external API is enabled, the desktop app exposes an MCP (Model Context Protocol) server at http://127.0.0.1:19528/mcp over Streamable HTTP. Any MCP-capable AI tool can connect with a one-line config and let the model list accounts, create publish tasks, and read logs directly.

Setup#

The Settings → External API section lets you copy this config with the port and token already filled in:

{
  "mcpServers": {
    "multipost-desktop": {
      "url": "http://127.0.0.1:19528/mcp",
      "headers": { "Authorization": "Bearer mp-xxxxxxxxxxxx" }
    }
  }
}

Claude Code#

claude mcp add --transport http multipost-desktop http://127.0.0.1:19528/mcp \
  --header "Authorization: Bearer mp-xxxxxxxxxxxx"

Cursor / other clients#

Write the JSON above into the tool's MCP config file (e.g. Cursor's ~/.cursor/mcp.json).

The token is the permission. Once it is in a config file, treat that file like a password. If the token leaks, regenerate it on the settings page.

Available tools#

Tool	Purpose
`list_accounts` / `get_account`	List / get accounts
`refresh_account`	Refresh login status and profile
`list_platforms`	List platforms and supported content types
`create_publish`	Create a publish task (async, returns groupId)
`get_publish_status`	Poll publish status
`list_publish_groups`	List open publish groups
`submit_publish` / `submit_publish_target`	Submit all / one target
`retry_publish_target`	Retry a failed target
`close_publish`	Close a publish group
`list_publish_history`	Query publish history
`list_views` / `get_group_data`	Inspect current views and group data
`execute_script` / `execute_group_script`	Run JS in a view
`get_console_logs` / `clear_console_logs` / `get_app_logs`	Read / clear logs
`take_screenshot`	Screenshot the main window or a publish target
`toggle_devtools`	Toggle DevTools for a publish target

create_publish is async: it returns a groupId immediately. Tell the model to then poll get_publish_status until all targets are terminal.

A typical flow#

list_accounts to get target account ids;
create_publish (autoSubmit: false) to create the task;
get_publish_status until targets become ready;
confirm, then submit_publish;
get_publish_status to verify success; on failure, retry_publish_target.

PreviousPublish NextDebug