# Connect & MCP

Source: https://magistermarketing.com/docs/mcp. Every page of this section, in order, as Markdown.

## How Connect works

Use Magister from your existing AI assistant, with your brand’s own hosted workspace.

**Connect** is the Magister plan for the AI you already use. You keep working in a compatible assistant such as Claude, ChatGPT, Codex, or another MCP client, and that assistant calls Magister's remote tools on your behalf.

**What Magister supplies** — each brand's own hosted workspace (its Library, files and memory), isolated brand profiles and assets, audits, full marketing plans, analytics, AI-visibility tracking, the marketing calendar, the Brief, integrations, publishing, social and paid-ad actions, and a shared permission history.

**What your assistant supplies** — the conversation, reasoning model, and client experience. Its model charges are billed by that provider and are separate from Magister's Connect credits.

**What Connect leaves out** — every Connect brand has its own hosted workspace, which sleeps when nobody needs it and wakes on demand. Magister's own agent is the Agent plan's: hosted chat, your own workflows on a schedule, skills, Slack, the hosted email inbox, browser control, and model choice are not part of Connect. The built-in workflows and the Brief run on Magister's side for every brand.

**One authorization, every brand** — one MCP connection covers every brand you can access in the organization you selected. Tools take a `project_id`, so your assistant can switch brands by using a different project from **List projects**. Reauthorize the connector to choose another organization.

For prices and allowances, see [Plans and pricing](https://magistermarketing.com/docs/billing/plans). For the canonical server URL, client commands, and complete tool list, use the [MCP connector hub](https://magistermarketing.com/mcp).

## Connecting your AI assistant

Authorize the remote MCP server, select an organization, approve scopes, and verify the connection.

Open the [MCP connector hub](https://magistermarketing.com/mcp) for the current server URL and copy-ready setup instructions. That page is the canonical reference for supported clients and tools.

**1. Add Magister** — add the remote Magister MCP URL in your assistant's connector or MCP settings. Claude Code and Codex commands, plus the Claude Desktop configuration, are available on the connector hub.

**2. Sign in and authorize** — your client opens Magister in a browser. Sign in, select the organization to connect, review the requested scopes, and approve. Read access is included by default; write, run, publish, send, spend, and media scopes are opt-in.

**Plan eligibility** — Magister checks the selected organization when you connect and on every MCP request. The current access policy may require an active Connect or Agent subscription. If the consent screen asks you to activate or fix billing, complete that step and then reconnect from your MCP client; checkout does not complete the earlier authorization request.

**3. Verify** — back in the assistant, ask it to "list my Magister projects." Pick a brand, then ask it to read the brand profile or current marketing plan.

**New account from an assistant** — the signup API on the connector hub can create a short-lived signup link after your assistant collects an email, team name, brand name, and website. Open that link and finish with Google sign-in; the account is created only after sign-in succeeds.

**Changing organizations** — one connection covers one selected organization. Reconnect or reauthorize in your MCP client's settings to choose a different organization; you do not need to reinstall the server.

**Managing clients** — the dashboard's Getting Started panel lists authorized MCP clients and their last-used time. Revoke an individual client there when you no longer trust or use it; other authorizations remain connected.

**Connection expired or missing tools** — reconnect from the client and approve the needed scope. If a tool needs a brand integration that is not connected, it returns a one-click Magister connection link.

## The Magister AI Visibility connector

A smaller Magister connector for Claude that only measures how a brand appears in AI answers.

**Magister AI Visibility** is a connector for Claude with one job: measuring how a brand appears in answers from AI engines such as ChatGPT, Claude and Gemini. It is an address of the same Magister server, with only the AI-visibility tools, and it asks for read access only.

**Address** — `https://mcp.magistermarketing.com/mcp/ai-visibility`. Add it from Claude's directory, or as a custom connector with that URL. Sign-in is the same browser sign-in as any other Magister connection, and it needs a Magister account with at least one brand.

**What it does** — lists your brands; reads a brand's prompt sets, scoring rounds and scores; reads the AI engine answers behind a score and the pages the engines cited; drafts prompts for a new set and saves the ones you confirm; starts a scoring round; starts an answer-engine (AEO) audit of your public website and reads its results; reads your credit balance.

**What it does not do** — anything outside AI visibility. It has no marketing plan, publishing, email, ads or connected-service tools, and it shows results as text rather than in-chat cards. For those, add Magister at its main address from the [MCP connector hub](https://magistermarketing.com/mcp). If you already added Magister at its main address, you do not need this connector: that connection includes everything here.

**Credits** — creating a set that runs its first round, starting a round, and starting an audit are metered against the organization's credits, exactly as they are elsewhere in Magister. Every read is free.

**A brand with no website** — the connector cannot add one. Open Magister, add the website there, and the first audit starts by itself.

## Connecting other apps' MCP servers

Give Magister tools it doesn't have built in, from any app that offers an MCP server.

This is the opposite direction from the rest of this section. Above, you add **Magister** to your AI assistant. Here, you add **another app's** MCP server to Magister, so your Magister agent can use that app.

**What an MCP server is** — an address an app publishes so an AI can use it: read its data and take actions through it. Linear, Asana, Atlassian (Jira and Confluence), Canva and many others offer one.

**Connecting one** — a brand owner or admin opens **Integrations → MCP servers** and picks a server, or chooses **Any MCP server** and pastes the address the app gave them, then signs in to the app and chooses what Magister may do with each of its tools. An owner or admin can also ask the agent, which sends them a link. Other members see the servers and ask an owner or admin to add one.

**You stay in control** — every tool that changes something asks you first until you allow it, unless the brand is on **Allow all**. A tool whose definition changes is held until you review it again. Magister keeps the sign-in on its servers, never on the agent's machine.

**Where they work** — your Magister agent, and AI tools connected at Magister's own MCP address, can use these servers. The ChatGPT and Claude apps added from their directories can't: their review rules don't allow a tool that runs other servers' tools.

**Limits** — Free includes 2 servers per brand; Agent and Connect include 20. Apps Magister already integrates with, such as Notion, HubSpot and GitHub, connect from **Discover** instead.

## Approvals, permissions, and safety

How OAuth scopes, immediate confirmations, category policy, and approval history work together.

Magister uses two layers of control: **connector scopes** decide which kinds of MCP tools are available, and the brand's **Permissions** policy decides which external actions must stop for review.

**Baseline access** — profile, plan, audit, AI-visibility, asset, social-analytics, and integration-discovery reads are available with the default scope. Built-in Magister research, audits, AI-visibility checks, plan recompiles, and media generation use that same baseline grant; metered calls remain bounded by the organization’s usage ledger and budget. External-account writes require their documented opt-in scope.

**Immediate confirmation** — publishing live content, sending email, spending money, deleting an asset, and other destructive actions require the relevant scope and an immediate user confirmation. MCP creates Meta, TikTok, and Google Ads Search campaigns atomically paused, and no others. Magister's own chat also creates Reddit campaigns, atomically paused, and LinkedIn, Pinterest, and X campaigns through immediate pause and readback, which can return an unresolved result that must be verified. Activating spend is a separate explicit action. Paid ads, live publishing, email send, and signed-in browser control remain separate permissions that even **Full access** does not silently include.

**Permissions** — open **Brand settings › Permissions** from the account menu. **Pending** shows requests waiting for a decision; **History** shows allowed, denied, running, successful, failed, and expired actions from Connect, Agent, workflows, and Slack. Open a request to review the exact destination, content, schedule, campaign, repository, or other details.

**Approve or deny** — approve to let the displayed action run, or deny it. A denial can include optional feedback for the agent, and the note remains in permission history rather than being sent to the external service.

**Returning to the assistant** — a high-risk MCP tool can return an operation ID and Magister review link. After you decide, the assistant polls the approval status and receives a redacted result rather than the raw action token or sensitive arguments.

**Standard permissions** — every new brand starts on Standard: media generation and routine data edits inside connected tools can run without asking, while email, social/content publishing, paid ads, code and pull requests, and destructive actions ask first. The selection is saved for the brand, so each new chat keeps the latest setting chosen by a brand admin; workflows, Slack, and Connect use it too.

**Ask-before categories** — brand admins can choose Ask first, Standard, Allow all, or their own category mix. Turning a category off allows matching actions to run without asking. Owners and admins can also choose **Always allow** on an approval card: it approves that request and turns that category's ask-before off for the brand, with Undo. Webflow publication and live Webflow changes always ask, whatever the setting; on **Allow all** they are the only requests that still ask. **Allow all** requires a strong confirmation; provider authorization, budgets, ownership checks, and other guardrails still apply. Paid spend and destructive approvals require a brand admin.

## Assets, integrations, and MCP tools

Work with generated brand documents, uploaded files, connected services, and the canonical tool catalog.

Connect uses the same brand Assets and integrations shown in the Magister dashboard.

**Read brand assets** — your assistant can list and read the live `BRAND.md` and `PLAN.md`, audit snapshots, archived plan versions, uploaded files, and generated media. Bounded text is returned directly; binary or oversized files use a one-hour signed download URL.

**Create assets** — with the `assets:write` scope, an assistant can add a UTF-8 text asset up to 1 MB or prepare and finalize an arbitrary file upload up to 50 MB. That same scope also lets it discover and snapshot selected hosted Agent text files from `resources/` into cloud Assets. A same-name upload creates a new item rather than overwriting an existing asset.

**Delete carefully** — uploaded assets can be deleted only with `assets:write` and immediate confirmation. Magister-generated brand, plan, and audit documents cannot be deleted through the MCP asset tool.

**Use integrations on demand** — list integrations to see connection state and supported execution. If a needed service is missing, the tool returns a one-click link. Credentials stay encrypted in Magister and are never exposed to the assistant.

**Canonical catalog** — tool names and scopes evolve faster than Help Center prose. Use the [full MCP tool reference](https://magistermarketing.com/mcp#tool-reference) for the authoritative current list.
