Developers

MCP server

Connect an AI agent to postme.live over the Model Context Protocol. The server mirrors the REST API one-to-one and authenticates with the same API key, no OAuth.

Already have an account? Issue an API key and follow the connect-your-agent quick start in Settings → Developers.

Open Settings → Developers

Endpoint

A remote, Streamable-HTTP MCP server. Point any MCP client at it and pass your API key as a bearer token.

text

https://postme.live/api/mcp

Authorization: Bearer pml_live_<lookupId>_<secret>

Every tool maps to an endpoint documented in the rest of this reference, so scopes, plan rate limits, idempotency, and audit logging behave identically. Mint a key at Settings → Developers.

Connect a client

Store the key in an environment variable and reference it. Most clients support ${env:VAR} interpolation, so the key never sits in plaintext in a config file.

Claude Code

bash

claude mcp add --transport http postme https://postme.live/api/mcp \
  --header "Authorization: Bearer $POSTME_API_KEY"

Cursor: ~/.cursor/mcp.json

json

{
  "mcpServers": {
    "postme": {
      "url": "https://postme.live/api/mcp",
      "headers": { "Authorization": "Bearer ${env:POSTME_API_KEY}" }
    }
  }
}

VS Code: .vscode/mcp.json

json

{
  "servers": {
    "postme": {
      "type": "http",
      "url": "https://postme.live/api/mcp",
      "headers": { "Authorization": "Bearer ${env:POSTME_API_KEY}" }
    }
  }
}

Windsurf and other clients that accept a remote MCP URL with custom headers work the same way: set serverUrl and an Authorization header.

Client support

Clients that let you set a bearer header on a remote server work today: Claude Code, Cursor, VS Code, and Windsurf.

ChatGPT and the Claude Desktop / claude.ai connector UI currently require OAuth for remote servers and cannot attach a static key, so they can't connect to this server yet. We'll add OAuth and a local (stdio) package if there's demand.

Tools

Twenty-two tools, one per API operation. A tool returns the same JSON the corresponding endpoint does, and a failure surfaces the API's structured error (e.g. a missing scope is the same 403 you'd get over REST).

Area	Tools
Identity	`postmelive_whoami`
Workspaces	`postmelive_list_workspaces`, `postmelive_get_workspace`
Brand	`postmelive_get_brand_kit`
Captions	`postmelive_tailor_captions`
Channels	`postmelive_list_channels`, `postmelive_get_channel`, `postmelive_get_channel_insights`
Media	`postmelive_list_media`, `postmelive_get_media`, `postmelive_create_media_upload`, `postmelive_upload_media`
Posts	`postmelive_list_posts`, `postmelive_get_post`, `postmelive_get_post_insights`, `postmelive_create_post`, `postmelive_update_post`, `postmelive_delete_post`
Suggested posts	`postmelive_list_suggested_posts`, `postmelive_approve_suggested_post`, `postmelive_create_video_card`, `postmelive_get_generation`

Uploading a local file takes three tool calls: postmelive_create_media_upload mints a presigned PUT URL, your agent sends the bytes to that URL with any HTTP client, then postmelive_upload_media registers the object by its key. A file that's already at a public HTTPS URL skips straight to postmelive_upload_media with a url — the server fetches it. The bytes never travel through the tool call itself.

Start a session with postmelive_whoami to discover the key's workspace, plan, and scopes.