Beta batch 4 is open: agent-drafted weekly reports now on Business. Request access
by Phronesis AI Lab
MCP server

Manage your social from Claude or ChatGPT

Connect Social Media Digital Marketer over the Model Context Protocol and run your whole workflow from a conversation. Pull performance, plan around what is working, draft on brand and send posts to your queue, with a person approving every publish.

What the assistant can do

Plan, draft and queue in a conversation

The same loop the app runs, reachable from your AI assistant. It proposes; a person decides.

Pull performance

Ask "what were my top posts this month?" and the assistant pulls real numbers from Social Media Digital Marketer: reach, reactions, comments, saves.

Plan the week

Turn what is working into next week's content ideas, matched to your brand model and the slots your channels actually perform in.

Draft on brand

Generate hooks, captions and variations in your voice. Edit a line in the conversation and regenerate only the part you would change.

Queue for approval

Content lands in your Social Media Digital Marketer queue, ready for a person to approve. The assistant proposes and drafts; it never publishes on its own.

Tool calls

Ask a question, get real numbers

The assistant calls the MCP tools it needs, reads the data back and turns it into a plan you can approve. The exchange below is illustrative.

  • Reads real performance from your workspace
  • Queues drafts, never publishes on its own
  • Scoped to the workspace the token connects
Social Media Digital Marketer · connected over MCP
What were my top posts this month, and draft one like the best.
tool call analytics.top_posts(period="30d")
↩ 3 posts · best: "We're hiring" · reach 3,589
tool call posts.draft(brief="hiring, founder voice")
Drafted a hiring post in your voice and added it to the queue as pending approval. A teammate approves before it publishes.

A human approves before anything publishes

The assistant can plan, draft and queue, but it cannot publish. Every draft waits for a person with the right role to approve it. You can approve a whole week at once; you cannot skip the approval.

Setup

Connect in three steps

From a token to a working conversation in a few minutes.

Get your MCP token

Open Social Media Digital Marketer, go to developer settings and create an MCP token for the workspace you want to connect. API and MCP access are included on Studio and above.

Add the server

In Claude or ChatGPT, add the MCP server with the URL and your token. The client discovers the available tools automatically.

Start a conversation

Ask about performance, plan the week or draft a post. Everything the assistant queues waits for a human approval before it publishes.

MCP FAQ

Questions about MCP

Which assistants can connect?

Any MCP-capable client. In practice that means Claude and ChatGPT today, with more clients as they adopt the Model Context Protocol. The Social Media Digital Marketer MCP server is in beta.

Does the assistant publish on its own?

No. The MCP server lets an assistant pull data, plan, draft and queue, but every draft waits for a person with the right role to approve it. Approval is a product rule you cannot skip.

Which plans include MCP?

API and MCP access are included on the Studio plan and above. See the plan matrix on the pricing page for the full breakdown.

What data can the assistant see?

Only the workspace the token is scoped to. It reads the performance and content it needs to plan and draft; it cannot reach other workspaces or publish without an approval.

How is this different from the API?

The API is for your own code. The MCP server is for AI assistants: it exposes the same actions as tools an assistant can call in a conversation, so you manage social by chatting instead of writing requests.

Run your social from a conversation

MCP access is included on Studio and above. A person still approves every publish.