Skip to content

Social Media MCP Server

A social media MCP server connects a compatible AI client to social-media operations through the Model Context Protocol. Once configured, an assistant such as Claude, Cursor, or ChatGPT can discover connected accounts, publish or schedule content, review analytics, and manage related records by calling exposed tools instead of requiring a separate local integration for each task. The server’s 34 tools span publishing, scheduling, analytics, integrations, profiles and AI helpers; the task table lists the common operations.

What an MCP server does for AI clients

Model Context Protocol is an open standard for letting AI assistants call external tools. An MCP server exposes those tools to a compatible client, so the assistant can request an operation, receive structured results, and continue the conversation with the outcome. For social-media work, that means account discovery, publishing, scheduling, post management, analytics, and platform-specific preparation can be handled through one client connection rather than through separate manual steps for every action.

Client configuration field map

Every client needs the MCP server URL and an access key. For Claude.ai web, go to Customize, then Connectors, then Add custom connector, and paste the URL into the Remote MCP field. Claude Desktop, Claude Code, and Cursor use JSON configuration with an mcpServers entry containing a url field and a headers field. ChatGPT and other Streamable-HTTP MCP clients use the same server URL and Authorization Bearer header, with the access_key query parameter as the fallback when custom headers are unavailable. After saving the configuration, restart the client and confirm that tools such as create_post, list_integrations, and get_analytics appear in the available-tools list.

Available MCP tools by task

Start with list_integrations when you need to confirm which social accounts are connected. Use create_post to publish or schedule content to one or more platforms. Use the post-management tools to browse, inspect, or delete posts, and use the scheduling tools to list, reschedule, or cancel queued posts. Use get_analytics for cross-platform post metrics. The AI helper tools support caption writing and image generation, while profile and app tools support multi-tenant profile management and BYOK OAuth app management. Pinterest and TikTok have dedicated prerequisite tools that should be called before create_post for those platforms.

Example prompts after connection

Once connected, test the connection with requests in this order: ask which social accounts are linked, then request a short announcement post queued for the next morning at 9am Eastern, then ask for performance figures covering the five most recent LinkedIn entries, and finally request artwork for a quarterly-results update to be shared on Twitter and LinkedIn. The assistant selects the matching tools, the operation runs, and the confirmation returns in the conversation. Begin with read-only discovery such as list_integrations before testing a publishing action.

Pinterest and TikTok prerequisite input contract

Pinterest requires a destination board. Call get_pinterest_boards, select a board from the returned results, and pass that board id as platformSpecificData.boardId on create_post. TikTok requires a privacy choice. Call get_tiktok_creator_info, inspect the returned privacyLevelOptions, and pass one of those values as platformSpecificData.privacyLevel on create_post. The create_post tool description points the model toward these prerequisites, but a careful operator should still verify that the prerequisite call happened and that the selected value came from the returned options rather than from memory.

MCP server authentication options

The MCP server accepts an access key in three forms, in this priority order: Authorization Bearer, x-access-key, and the access_key query parameter. The Bearer form is the recommended option when the client supports custom headers. The x-access-key form uses the same header name as the REST API. The query-string form is reserved for clients without custom-header support. The MCP server forwards calls for authentication handling, so a key that does not work against the REST API should not be expected to work through the MCP connection. The MCP server does not validate the key itself: it forwards every call to the REST API, which enforces authentication, credit usage and rate limits. Plan limits and access-key permissions therefore behave identically through MCP and REST.

Self-hosting requirements and limits

Self-hosting is an option when an MCP server must run on owned infrastructure for compliance, a custom domain, or another operational requirement. The stated runtime requirements are Node, Express, and @modelcontextprotocol/sdk, and the server can run anywhere Node runs, including a VPS or a hosted Node platform. Self-hosting does not remove the need to configure authentication, verify tool discovery in the client, or test the Pinterest and TikTok prerequisite flows before relying on publishing actions.

Troubleshooting failure matrix

A 401 Unauthorized response means the access key is missing or invalid, so confirm that the Bearer token is present in the client configuration and that the same key works against the REST API. A 429 Too Many Requests response means the applicable rate limit has been reached, so either upgrade or space the calls out. If tools do not appear, restart the AI client after editing the configuration and check whether MCP is enabled explicitly in the client settings.

Client configuration field map
Client typeMCP URL fieldAuthentication fieldFallback
Claude.ai webCustomize > Connectors > Add custom connector; Remote MCP URL: {{MCP_BASE_URL}}/mcp?access_key=YOUR_API_KEYAccess key embedded in the remote MCP URLNot applicable
Claude Desktop, Claude Code, and CursormcpServers entry url: {{MCP_BASE_URL}}/mcpheaders.Authorization: Bearer YOUR_API_KEYNot stated for this JSON example
ChatGPT and other Streamable-HTTP MCP clientsServer URL: {{MCP_BASE_URL}}/mcpAuthorization: Bearer YOUR_API_KEYaccess_key query parameter when custom headers are unsupported
Other header-capable clients{{MCP_BASE_URL}}/mcpx-access-key: YOUR_API_KEYaccess_key query parameter for clients without custom headers

Use the MCP origin consistently; the server accepts the authentication methods in the priority described below.

Available MCP tools by task
Task groupToolsUse
Account discoverylist_integrationsShow connected social accounts
Publishingcreate_postPublish or schedule a post to one or more platforms
Post managementlist_posts, get_post, delete_postBrowse and manage posts
Schedulinglist_scheduled_posts, reschedule_post, cancel_scheduled_postManage the publish queue
Analyticsget_analyticsRetrieve cross-platform post metrics
AI helpersai_write_content, ai_generate_imageSupport captions and images
Platform prerequisitesget_pinterest_boards, get_tiktok_creator_infoPrepare Pinterest and TikTok publishing inputs
Profilescreate_profile, list_profiles, update_profileManage multi-tenant profiles
Appscreate_app, list_appsManage BYOK OAuth apps

Common operations from the 34-tool server inventory. Full tool schemas are discovered by the client.

Pinterest and TikTok prerequisite input contract
PlatformPrerequisite toolField to pass on create_postRequired rule
Pinterestget_pinterest_boardsplatformSpecificData.boardIdSelect a returned board id; pins must land on a board
TikTokget_tiktok_creator_infoplatformSpecificData.privacyLevelUse one of the returned privacyLevelOptions values

Exact field names preserved from the source contract; returned option values must come from the prerequisite tool response.

Troubleshooting failure matrix
SymptomSource-stated causeCheck
401 UnauthorizedAccess key is missing or invalidConfirm the Bearer token in the client config and verify the key against the REST API
429 Too Many RequestsApplicable rate limit reachedUpgrade or space the calls out
Tools do not appear in the clientClient has not reloaded the edited configuration or MCP is not enabledRestart the AI client and confirm MCP is enabled in settings

Only source-stated symptoms and checks are included; no new error codes or remedies are introduced.

Frequently asked questions

Which authentication method should I use for an MCP server?

Use the Authorization Bearer header when the client supports custom headers. Use x-access-key when that header is the expected form. Use the access_key query parameter only when the client does not support custom headers.

What must happen before posting to Pinterest or TikTok through create_post?

Call get_pinterest_boards first, choose a board, and pass its id as platformSpecificData.boardId. Call get_tiktok_creator_info first and pass one of the returned privacyLevelOptions values as platformSpecificData.privacyLevel.

Why do MCP tools not appear after I update the client configuration?

Restart the AI client after editing its MCP configuration. Some clients also require MCP to be enabled explicitly in settings before the tools appear.

Claude.ai web remote MCP URL example
{{MCP_BASE_URL}}/mcp?access_key=YOUR_API_KEY
Claude Desktop, Claude Code, and Cursor MCP JSON configuration example
{
  "mcpServers": {
    "social-media": {
      "url": "{{MCP_BASE_URL}}/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}
Claude.ai web remote MCP URL
{{MCP_BASE_URL}}/mcp?access_key=YOUR_API_KEY
Claude Desktop, Claude Code, and Cursor MCP configuration
{
	"mcpServers": {
		"{{MCP_SERVER_NAME}}": {
			"url": "{{MCP_BASE_URL}}/mcp",
			"headers": {
				"Authorization": "Bearer YOUR_API_KEY"
			}
		}
	}
}