Let your AI agent handle the posting.

Connect your AI assistant to Content Mill and ask for posts in plain words. The agent drafts them, checks each platform’s limits and schedules them, using the same rules and credits as the app. There is no API key to copy: it signs in the way you do.

Connect your agent What it can and can’t do

Claude
ChatGPT
Cursor
Claude Code
Codex
VS Code
any MCP client

You need a Content Mill account first. Create one or sign in.

Your AI agentExample

Schedule the launch post for Friday 9am on X and LinkedIn.
list_social_accounts2 accounts connected
get_platform_limitsX 280 · LinkedIn 3,000
create_draftDraft saved
preview_postPreview ready, nothing sent
Here is the preview: Friday 9:00 AM on X and LinkedIn. Shall I schedule it?
Yes, schedule it.
publish_postScheduled for Friday 9:00 AM

From a chat to your accounts

Step 1: You ask
In Claude, ChatGPT or your editor, in plain words.
Step 2: The agent works
It looks up your accounts and limits and saves drafts, through Content Mill’s tools.
Step 3: A preview comes first
Nothing can be published until a preview has been made.
Step 4: It sends exactly that
The post that goes out is the one the preview showed, once.

Connect your agent

Use this address whenever an assistant asks for a remote MCP server, then pick yours below.

Server address

https://mcp.contentmill.dev/mcp

Claude

Open Settings, then Connectors, then Add custom connector.
Name it Content Mill and paste the server address above.
Click Connect. Sign in with your Content Mill email and password, or Google, then approve access on the consent screen.

ChatGPT

Open Settings, then Apps, then Advanced settings, and turn on developer mode.
Create an app, choose MCP server, and paste the server address above.
Choose OAuth when asked how to authenticate. Sign in with your Content Mill email and password, or Google, then approve access on the consent screen.

Claude Code

Run this command in a terminal.

Terminal

claude mcp add --transport http content-mill https://mcp.contentmill.dev/mcp

Start a session and run /mcp, or use a Content Mill tool. Claude Code opens the sign-in page in your browser.
Sign in with your Content Mill email and password, or Google, then approve access on the consent screen.

Codex

Run this command in a terminal.

Terminal

codex mcp add content-mill --url https://mcp.contentmill.dev/mcp

Codex prints a sign-in URL. Open it in your browser.
Sign in with your Content Mill email and password, or Google, then approve access on the consent screen.

Cursor

1
Add this to ~/.cursor/mcp.json, or to .cursor/mcp.json inside a single project.
~/.cursor/mcp.json
```
{
  "mcpServers": {
    "content-mill": {
      "url": "https://mcp.contentmill.dev/mcp"
    }
  }
}
```
Run cursor-agent mcp login content-mill in a terminal, or open a chat in Cursor and use a tool. Either one prompts you to sign in.
Sign in with your Content Mill email and password, or Google, then approve access on the consent screen.

VS Code

1
Open the Command Palette and run MCP: Add Server, choose HTTP, and paste the server address above. Or add it directly to .vscode/mcp.json:
.vscode/mcp.json
```
{
  "servers": {
    "content-mill": {
      "url": "https://mcp.contentmill.dev/mcp",
      "type": "http"
    }
  }
}
```
Start a chat and use a Content Mill tool. VS Code opens the sign-in page in your browser.
Sign in with your Content Mill email and password, or Google, then approve access on the consent screen.

Any other MCP client

Sign-in is OAuth 2.1 with PKCE. Clients can identify themselves with a Client ID Metadata Document or register themselves dynamically, so most MCP clients connect without anything being set up on this side first.

Open the assistant’s settings and look for Connectors, Plugins or MCP servers.
Add a custom or remote server with the address above, and choose OAuth as the authentication method.
Sign in and approve access when it asks.

What your agent can and can’t do

These limits are enforced by Content Mill itself, whatever the assistant tries.

It can

Read your connected accounts and your credit balance
Check each platform’s limits before it writes
Save drafts, edit them, and delete them
Prepare a post for your approval
Upload an image or video from a public link
Look up what is scheduled and what went out
Cancel a post that is still scheduled

It can’t

Publish a post it has not previewed first
Publish anything other than the exact preview — change a word and it needs a new one
Reuse a preview: each is single-use and expires in 10 minutes
Edit a scheduled or published post
See your passwords or the tokens behind your accounts

The assistant is instructed to show you the preview and wait for your yes before it publishes. That yes is given in your conversation; Content Mill guarantees the preview step and keeps a record of every call.

The tools your agent gets

What the assistant sees once it is connected.

Look things up

Read-only and free.

list_social_accounts: Your connected accounts and whether each is connected right now.
get_credit_balance: Your balance and what a post costs.
get_platform_limits: Character, media and format limits for the platforms you name.
list_posts: Your posts and their status.
get_post: One post’s result on each platform, with the link to the live post.

Prepare

Nothing goes public.

upload_media_from_url: Fetch an image or video from a public link and store it for posts.
create_media_upload: Reserve a one-time slot for a file with no link, such as a chat attachment.
finish_media_upload: Store the file that was posted to that slot, with the same checks as a web upload.
create_draft: Save a post as a draft, checked against each platform’s limits.
update_draft: Replace a draft’s contents (the old version isn’t kept). Only drafts can be edited.
delete_post: Delete a draft, or cancel a post that is still scheduled.

Publish

Two steps, in order.

preview_post: Validate and price a post, and return a single-use preview token. Publishes nothing.
publish_post: Publish or schedule exactly the post that preview showed, once.

Files from a chat

An assistant can’t pass a file through a tool call; it would have to retype every byte, which is slow and unreliable. So there is no base64 option. If the file has a public link, the assistant uses that. If it doesn’t, it reserves a one-time upload slot, posts the file to it from its own shell or HTTP tool, and then claims it. The slot works once and expires after 10 minutes, and files are limited to 50MB.

This needs an assistant that can run a command or make a web request. One that can only see the attachment can’t use it; in that case give it a public link, or add the media in the web app.

Create your account