Skip to main content

​
Working with her in the terminal

Claude Code is Anthropic’s agentic coding tool. It lives in your terminal, runs real commands and speaks the Model Context Protocol, which is everything she needs to join the session you already have open. You finish the feature, you say what you want the world to see, and the announcement is on your calendar before you have switched windows.
Claude Code in a terminal: you ask in plain words and it runs the postqueen CLI to schedule the launch posts

​
Set up

Two routes lead to the same place. Her skill teaches Claude Code the postqueen CLI, which is the one to pick when you want shell commands and JSON you can pipe into jq. MCP hands it her 11 hosted tools instead, with nothing installed locally. Running both is fine.
1

Install the CLI

2

Install her skill

What lands is her SKILL.md, registered under the name postqueen and declaring Bash(postqueen:*) as the only tool it needs. Claude Code picks it up on its own when a task looks like posting work, so it arrives already knowing the discovery flow, the flags on posts:create and the rule about media.
3

Export your API key

Your key lives in the app at app.postqueen.ai: open Settings > Developers > Public API and click Reveal. Export it in the shell you launch Claude Code from:
Self-hosting PostQueen? Point the CLI at your backend URL, the same value as NEXT_PUBLIC_BACKEND_URL. Most single-domain setups serve the backend under /api:
4

Verify the connection

A JSON list of your connected channels means she is ready. If a command starts failing later, postqueen auth:status shows which credentials the CLI is using, though it does not contact the server, so this same command is the real test.
5

Or connect over MCP

Prefer tool calls to shell commands? It is one claude mcp add line, and the scoping options, the header variant and self-hosted URLs are all on Claude Code over MCP.
postqueen auth:login runs an OAuth2 device flow against a self-hosted auth server, and there is no hosted one at the moment. On the hosted service, the API key is the route to take.

​
👑 A launch post with a screenshot

You just shipped dark mode and grabbed a screenshot of it. Ask in plain words:
“Schedule a launch post about dark mode for tomorrow at 9am on X and LinkedIn, with screenshots/dark-mode.png attached.”
It starts by finding your channels and their integration IDs:
The screenshot goes through upload next, because -m takes a PostQueen URL rather than a local path. Media has to live on a PostQueen domain, and the API rejects the request otherwise. The tail -n +2 drops the command’s header line, which is not JSON:
Then it schedules the post. The -s flag is required and takes an ISO 8601 date, and -i takes the integration IDs from the first command. X requires who_can_reply_post, which LinkedIn ignores, so one call still covers both channels:
The per-provider requirements are listed in X provider settings and its neighbors. Finally it reads back what it queued, so you get a confirmation instead of a promise:
The post now sits on your calendar at app.postqueen.ai. She runs on your terms: leave it there and she publishes it at nine tomorrow, on time, without another word from you. Want to look first? Add -t draft to the same command and the post lands on the calendar unscheduled, waiting until you publish it.
Over MCP the conversation is the same and only the plumbing changes. Claude Code calls integrationList, brings media in with uploadFromUrlTool when it has a public URL, then schedules with integrationSchedulePostTool. Each tool is documented in the MCP tool reference.

​
FAQ

No. She works on her own through the dashboard, the CLI and the public API. Claude Code is one of several ways to automate her, and the one to pick when your day already happens in a terminal.
Whichever fits the way you work. The skill plus the CLI gives you shell commands with JSON output you can pipe and script, which suits repo work and cron jobs. MCP gives Claude Code 11 tools with no local install, which suits a machine where you would rather not add a global npm package. The two coexist happily, so you can start with one and add the other later.
Everything the CLI can do. It lists your channels and reads what each one accepts, creates scheduled posts or drafts with media attached, and pulls analytics once they are live. Commands print a one-line header and then JSON, so the agent reads a field rather than guessing at wording on screen. Pipe through tail -n +2 before jq.
Yes. Ask for drafts and each post lands on the calendar unscheduled, waiting until you publish it there. Anything already scheduled is visible on the same calendar before it goes out, so you can still edit or delete it.

​
Learn more

CLI introduction

The postqueen CLI that Claude Code drives, with installation and the full command set.

Claude Code over MCP

Project config, the Authorization-header variant and self-hosted URLs.

Why agents love her

Why an agent can drive her without guesswork, and the two rules that matter most.

Agents overview

Every way to put an agent in front of PostQueen, from coding assistants to chat apps.