What access does Advino need to my clients' accounts?

Read access for monitoring, plus the pause permission if you want auto-pause armed. Connection is standard Meta OAuth — no passwords, no billing credentials, revocable from Business Manager any time. Disconnecting a client takes one click on either side.

Is auto-pause safe? I don't want a robot touching live campaigns.

Auto-pause is off by default and only ever follows rules you write — for example, "pause ad sets spending over $200/day if the payment method fails." Every action is logged with a timestamp and a one-click undo. Plenty of agencies run alert-only and pause manually; that's a fine way to start.

Does this replace my reporting dashboard?

No, and it isn't trying to. Dashboards answer questions when you look at them. Advino is the layer underneath that's watching when you're not looking — nights, weekends, the Tuesday you're stuck in onboarding calls. Keep your dashboard. Advino covers the hours it can't.

Meta now has its own alerts and an official Ads MCP server. Why pay for this?

Meta's notifications are tuned to protect Meta's revenue, not yours — slow, easy to miss, and silent on the failures that matter most to agencies. Meta's official Ads MCP (mcp.facebook.com/ads) is powerful, but it's single-advertiser: it manages and reports on one ad account at a time, on demand. Advino works one level up — it watches your entire book continuously (a 15-minute cycle, your thresholds, alerts on channels you actually check, action when it counts) and answers cross-account questions like spend by client and supplier that a per-account tool structurally can't. Use both: Meta's MCP for one account, Advino for the whole agency.

How long does setup take, honestly?

Connecting your first account is about two minutes of OAuth clicks. Sane default tripwires are on from the first scan; tightening thresholds per client takes another ten if you bother (most don't until the first alert fires). No calls required at any point.

For Claude · ChatGPT · Claude Code · Cursor

Ask Claude across your whole book of accounts.

The Advino MCP server is a Model Context Protocol endpoint that lets Claude or ChatGPT answer across every client and supplier at once — total spend by client, which accounts are disabled right now, true ROAS reconciled against Shopify, and creative scores — in plain English. Meta's official MCP manages one advertiser at a time; Advino answers for the entire agency book, and tells you whether the accounts are safe and what they really earned.

Works alongside Meta's official MCP (mcp.facebook.com/ads) — connect both in the same session.

What is MCP?

Model Context Protocol is an open standard for letting AI assistants talk to external systems. Anthropic released it in 2024. Claude, ChatGPT, Cursor, and most agent frameworks now speak it.

Meta shipped an MCP server for the Marketing API on April 29, 2026. Their server answers questions like "what did this campaign spend yesterday?" Advino's server answers different questions: "which accounts have an active risk signal?" or "what's the real ROAS gap vs Shopify?"

Sample prompts

Ask in plain English.

"How much did every client spend last week, by supplier?"

list_clientsaccount_spend

"What's my account health across all clients?"

get_account_health

"Show creatives scored below 40 this week."

get_creative_scores

"When was the last Shield auto-pause and why?"

get_alert_historylist_active_shields

"What's my real ROAS vs Meta-reported for Atlas this month?"

get_true_roas

What's available

Plan gating.

Read tools have daily quotas on lower tiers. Write tools require human approval in the Advino app — Claude never silently pauses live campaigns.

Tool	Type	Trial	Solo $29	Studio $79	Agency $199	Scale $499
get_account_health	read	10/day	100/day	∞	∞	∞
get_alert_history	read	24h	30d	∞	∞	∞
list_active_shields	read	✓	✓	✓	✓	✓
get_creative_scores	read	—	1 acct	5	15	40
get_true_roas	read	—	—	✓	✓	✓
get_shield_rules	read	—	✓	✓	✓	✓
create_alert_rule	write	—	—	✓	✓	✓
pause_ad_set	write	—	—	approval	approval	approval
resume_ad_set	write	—	—	approval	approval	approval
bulk_action	write	—	—	—	approval	approval

FAQ

The Advino MCP, answered.

How is the Advino MCP different from Meta's official MCP server?

Meta's official Ads MCP (launched 2026, at mcp.facebook.com/ads) is powerful but single-advertiser: it manages and reports on one ad account at a time. The Advino MCP works one level up — across your whole agency book. Ask one question and get an answer spanning every client and supplier at once: total spend by client, which accounts are disabled right now, real ROAS reconciled against the store, creative scores. It's the cross-account, monitoring layer a per-account tool structurally can't provide — and you can connect both MCPs in the same session.

What can I ask the Advino MCP in Claude or ChatGPT?

Book-wide questions no single-account tool can answer, such as: how much every client spent last week broken down by supplier, which client accounts have an active risk signal right now, what your real ROAS is versus Meta-reported, which creatives scored below a threshold this week, and when the last Account Shield auto-pause fired and why.

Can the MCP pause or change my live campaigns?

Write actions (like pausing an ad set) require explicit human approval inside the Advino app — Claude or ChatGPT never silently changes live campaigns. Read tools are quota-limited on lower tiers; write tools are approval-gated on Studio and above.

Which AI clients work with the Advino MCP?

Any Model Context Protocol client: Claude (desktop and web), ChatGPT, Claude Code, and Cursor. The same endpoint (mcp.advino.app) and token format work across all of them.

Install

About a minute.

Generate a token
Sign in to Advino, go to Settings → MCP, click "Generate token." Advino builds the exact snippet for your client with the token already baked in — copy, paste, done. Prefer to wire it up by hand? Steps 2–3 below.

Add the server to your client

Claude Code — one command, no files to edit:

claude mcp add --transport sse advino https://mcp.advino.app/sse \
  --header "Authorization: Bearer adv_mcp_live_<your-token>"

Claude Desktop — add this to ~/Library/Application Support/Claude/claude_desktop_config.json. This uses mcp-remote, which works on every version (needs Node/npx):

{
  "mcpServers": {
    "advino": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.advino.app/sse",
        "--header",
        "Authorization: Bearer adv_mcp_live_<your-token>"
      ]
    }
  }
}

ChatGPT & Cursor (and newer Claude Desktop builds) accept the remote URL directly — paste the endpoint https://mcp.advino.app/sse and the Authorization: Bearer … header into the client's connector panel, or for Cursor into ~/.cursor/mcp.json:

{
  "mcpServers": {
    "advino": {
      "url": "https://mcp.advino.app/sse",
      "headers": { "Authorization": "Bearer adv_mcp_live_<your-token>" }
    }
  }
}

Restart and query
Restart the client. You should see advino in the MCP server list. Try the prompt "What's my account health?" — it should call get_account_health.

One token, any client. The token authorizes the agency, not the specific app — use one across Claude, ChatGPT, and Cursor, or generate one per device for auditability. Full walkthrough: Connect your Meta ad accounts to Claude in 60 seconds.