Connect an AI assistant (MCP)

MCP is how AI assistants plug into other apps. Connect Publiz and you can ask Claude things like “draft a LinkedIn post about our new menu and schedule it for Monday at 9” or “how many AI credits do we have left?”, and it does it in your workspace.

The assistant works through an API key, so it can do exactly what that key can do and nothing else. It acts in one workspace, the one the key belongs to.

Connect it

Create an API key under Developer then API Keys. Tick only the scopes you want the assistant to have. A key with Posts set to Read only can look at your posts but never change them, which is a nice way to start.
Add Publiz to your assistant. In Claude Code:
Terminal window
```
claude mcp add --transport http publiz https://mcp.publiz.app/mcp \
  --header "X-API-Key: publiz_sk_your_key_here"
```
Using Claude Desktop or claude.ai instead? See below.

Other clients that take a remote MCP server want the same two things: the URL https://mcp.publiz.app/mcp and the header X-API-Key: publiz_sk_your_key_here.
Ask it something. “Which Publiz workspace are you connected to?” is a good first check.

Claude Desktop and claude.ai

Add Publiz once as a custom connector and it shows up in both.

In Claude, open Settings then Connectors and press Add custom connector.
Name it Publiz and paste https://mcp.publiz.app/mcp as the URL.
Under Authentication, pick No sign-in. Claude may pre-select Sign in now, but Publiz uses an API key, not a sign-in.
Under Request headers, add a custom header named X-API-Key with your key as the value, just publiz_sk_… with nothing in front.
Press Add. Leave the transport on Streamable HTTP.

Claude asks before every tool call at first. If that gets tiring, open the connector and set Read-only tools to always allowed, and keep approval on for anything that writes or deletes.

Working across two workspaces? Make a key in each and add Publiz twice under different names, like publiz-northbrew and publiz-agency.

What it can do

The assistant only sees the tools your key can use. Here’s the full set:

Scope	What the assistant can do
Always	Tell you which workspace it’s in and what the key can reach
Posts	List and read posts, create drafts or scheduled posts, edit, delete, publish now, retry a failed post. Posts can carry tags, the idea they came from, and a different signature (or none) per network
Tags	List, create, rename and delete post tags
Signatures	List, write, edit and delete signatures
Personas	List, read, create, edit, duplicate and delete personas
Usage	Check plan usage, AI credits, X credits and AI spend history
Ideas	List, save, delete and generate ideas (in a persona’s voice if you like), turn an idea into a draft, and set its feature image
Media	List and look up assets in the media library
Accounts or Channels	List the connected social accounts and their connection health
Analytics	Workspace totals and per-post numbers
Insights	The best times to post
Automation	List, write, edit and delete the reply templates comment rules send
AI	Every AI tool (images, video, voice, carousels and more) as its own tool, with a cost estimate before it runs, and checking on the job. It can also write a post’s first comment, in a persona’s voice if you like, and then save it on the post with Posts

Which scopes a key can hold depends on your plan, see the table on the API keys page. On Free that’s Posts, Usage and Signatures.

Connecting a new social account, uploading files and the comment tools stay in the dashboard for now.

Limits

Every tool call is an API request, so it counts toward your plan’s daily API requests (100 a day on Free). The assistant also checks what your key can do when it loads its tools, which costs one request now and then.

When something isn’t allowed, the assistant gets Publiz’s reason word for word, like “the AI scope is not enabled on your plan”, and can tell you what to change.

Turning it off

Revoke or edit the key under Developer then API Keys. The assistant loses access straight away.