MCP Server

AdBliss for AI Tools

Connect Claude, Cursor, and other AI tools directly to your ad performance data, revenue metrics, and platform connectors — no API calls to write, no dashboards to copy-paste.

Overview

The AdBliss MCP server runs at https://mcp.adbliss.io/mcp and implements the Model Context Protocol. Once connected, your AI tool can query campaigns, pull revenue data, model the impact of budget changes, and fire off real-time platform API calls — all from chat.

28 tools

Exploration, actions, simulation, live API

17 platforms

Meta, Google, Stripe, Shopify, and more

Two auth modes

API key or OAuth — your choice

The MCP server is separate from the REST API. You don't need to write any API calls — your AI tool sends tool invocations directly. For the REST API, see the API reference.

Requirements

  • •An AdBliss account with at least one connected platform — set one up at Settings → Connections
  • •A compatible AI tool: Claude Code, Claude Desktop, Cursor, Gemini CLI, or VS Code with Copilot Agent mode

Connect to the MCP server

AdBliss uses OAuth to authenticate MCP connections — the same standard all MCP clients support by default. Point your tool at the server URL, and it will walk you through a one-time login in your browser. That's it — no API keys to manage, and org selection is handled during the OAuth flow.

The first time you connect, your AI tool will open a browser window to authorize AdBliss. After that, the session is persistent and no further steps are needed.

Claude Code

Recommended

Step 1 — Add the server

terminal
claude mcp add adbliss --transport http https://mcp.adbliss.io/mcp

Step 2 — Authorize

On first use, Claude Code will prompt you to open a browser link. Log in with your AdBliss account to complete authorization.

Step 3 — Try it

chat
What's my ROAS across all platforms this week?

Claude Desktop

Add the following to your claude_desktop_config.json (found at ~/Library/Application Support/Claude/ on macOS):

json
{
  "mcpServers": {
    "adbliss": {
      "url": "https://mcp.adbliss.io/mcp"
    }
  }
}

Restart Claude Desktop. On first tool use it will prompt you to authorize via browser.


Cursor

Add to .cursor/mcp.json in your project root, or ~/.cursor/mcp.json for global access:

json
{
  "mcpServers": {
    "adbliss": {
      "url": "https://mcp.adbliss.io/mcp"
    }
  }
}

Restart Cursor. AdBliss tools appear in the Composer agent tool list after you authorize.


Gemini CLI

terminal
gemini mcp add --name adbliss --url https://mcp.adbliss.io/mcp

VS Code

Requires Copilot Agent mode. Add to .vscode/mcp.json in your workspace:

json
{
  "servers": {
    "adbliss": {
      "type": "http",
      "url": "https://mcp.adbliss.io/mcp"
    }
  }
}

Open the Command Palette (Cmd+Shift+P) and run MCP: Refresh Servers.


Exploration tools

Read-only queries against your synced data. Each call counts toward your monthly tool quota.

ping

Health check — verifies connectivity and returns your org ID and a server timestamp.

list_active_connectors

Lists all connected platforms for your org. Filter by type (ad_platform, payments, ecommerce, crm, etc.) and optionally include data freshness stats.

get_portfolio_snapshot

Full cross-platform overview: all campaigns, top ad groups, and platform totals for a configurable date range (1–90 days).

query_ad_metrics

Query performance, child entities, creatives, audiences, or budgets for any campaign, ad group, or ad across Meta and Google Ads.

get_entity_detail

Deep-dive on a single campaign or ad group — definition, budget, daily time series, and child entity performance.

explain_change

Why a metric moved between two windows — the change attributed to platforms, campaigns or ad sets with steps that sum exactly, split into own-rate vs mix for ratios, plus the ROAS/CAC rate chain.

query_revenue

Revenue from Stripe, Shopify, and WooCommerce, grouped by day, week, month, or platform. There is no CAC scoping or authorized/settled timing input — use query_growth scope=cac for acquisition-scoped analysis.

query_growth

CAC trends with power-law projections, or saturation signals — rising CPM/CPC alongside falling CTR.

query_unified_data

Cross-platform metrics grouped by day, platform, or campaign (campaign is ad-side only). Unsupported grains are refused by name, never answered at a different grain.

query_contacts

CRM deal pipeline by stage, or communications engagement (email, SMS, push) from connected marketing tools.

get_changes_since_last_run

Detects budget drifts, status changes, and metric shifts greater than 25% since the last analysis run.

calculate

In-session arithmetic: budget change amounts, percent differences, ROAS analysis, side-by-side entity comparison, and marginal efficiency modeling.


Simulation tools

Model the impact of a change before recommending it. Uses a power-law model calibrated to your historical data.

simulate_change

Projects the impact of pausing, enabling, increasing/decreasing budget, reallocating spend, changing bids, or adjusting schedules. Returns projected CAC and conversion impact with a fit_status indicator.


Live API tools

Direct, uncached queries to connected platforms in real time. Use these when you need data fresher than the last sync cycle.

query_api

Execute a real-time API call to a connected platform. Set connector to one of the live platform keys (meta_ads, google_ads, stripe, shopify) and endpoint to the operation you want. See Live API Endpoints for the full list.


Action tools

Write recommendations to your AdBliss dashboard for review. These tools do not make changes to your ad platforms directly — they queue proposals in your recommendations inbox.

set_budget

Recommend a daily or lifetime budget change for a campaign or ad group.

set_status

Recommend pausing or enabling a campaign, ad group, ad set, or ad.

set_bid

Recommend a bid strategy or amount change.

reallocate_budget

Recommend moving spend from one entity to another.

set_audience

Recommend an audience targeting change.

set_schedule

Recommend a dayparting or flight schedule change.

compound_action

Bundle multiple recommendations into a single atomic proposal.

Recommendations land in Dashboard → Recommendations. Your team reviews and approves them before any platform change is applied.

Authorization tools

connector_authorize

Start the OAuth flow for a new ad-platform connection. Returns a short URL — open it in your browser to complete authorization. OAuth today: google (Google Ads), facebook (Meta Ads). Stripe connects via a restricted-key paste in the dashboard; Shopify via the sidecar install link. Other platforms are on the roadmap.

list_authorized_orgs

Returns the organizations your current session is authorized for. Unmetered — does not count against your monthly quota.

Connecting a new platform from chat

chat
Connect our Google Ads account.

The AI will call connector_authorize, return a short link, and wait for you to complete the OAuth flow in your browser. Once authorized, the connector is live immediately. connector_authorize handles Google Ads and Meta Ads only — Stripe and Shopify are connected by a human in the dashboard.


Supported platforms

Connect platforms from Settings → Connections, or use connector_authorize from chat for the two ad platforms it supports. The list_available_connectors tool returns this same split as JSON.

Available now — ad platforms

Meta (Facebook Ads)OAuth
Google AdsOAuth
The Trade DeskAPI token + My Reports template

Available now — payments & e-commerce

StripeAPI key
ShopifySidecar install link
WooCommerceAPI key

Available now — analytics

Google Analytics 4OAuth

Wired, awaiting app registration

Pinterest AdsOAuth (app not yet registered)

Roadmap — not connectable today

TikTok AdsNot available yet
LinkedIn AdsNot available yet
HubSpotNot available yet
KlaviyoNot available yet
AttentiveNot available yet
MailchimpNot available yet
PaddleNot available yet
ChargebeeNot available yet
RecurlyNot available yet
Lemon SqueezyNot available yet
JobberNot available yet

Live API endpoints

Use these with the query_api tool. Set connector to the platform key and endpoint to one of the values below.

meta_ads

campaign_detailscampaign_performancead_set_detailsad_detailsad_creativesaudience_breakdowndevice_performanceplacement_performancegeographic_performanceconversion_actionschange_historyclient_accountssaved_audiences

google_ads

campaign_detailskeyword_performancesearch_termsaudience_breakdowndevice_performancegeographic_performanceclient_accountscampaign_metadataad_group_detailsad_detailsad_creativesconversion_actionschange_history

stripe

recent_transactionssubscription_detailscustomer_detailsrefundsproduct_performancebalance_transactionpayment_intentdispute

shopify

recent_transactionssubscription_detailscustomer_detailsrefundsproduct_performance

Example prompts

Once connected, you can ask your AI tool questions like these directly from chat.

“What's our ROAS across all platforms for the last 30 days?”
“Which campaigns have the highest CAC this month? Show me the trend over 90 days.”
“Simulate what happens to our Meta ROAS if we increase budget on campaign 123456 by 20%.”
“Connect our Shopify store — the domain is mystore.myshopify.com.”
“Show me our top 5 ad groups by spend this week and flag any that look saturated.”
“Pull the latest keyword performance from Google Ads for our brand campaigns.”
“What changed since the last analysis run? Any budget drifts or sudden metric drops?”

Troubleshooting

Authorization prompt doesn't appear

Make sure your browser isn't blocking pop-ups from mcp.adbliss.io. Try opening the auth URL manually if your AI tool prints it to the terminal.

Unknown tool after a server update

Tool schemas are cached at session start. Restart your AI tool session or toggle the MCP server off and on in your settings.

403 MCP not enabled

MCP access is controlled per user. Contact your account admin to enable MCP access for your user.

429 Too Many Requests

You've hit the monthly tool quota. Upgrade your plan at Settings → Billing, or wait for the quota to reset at the start of next month.

Tools appear after deploy but return stale data

Durable Object instances take 30–60 seconds to pick up new server code after a deploy. Wait a minute and retry.

Ready to connect?

Create an account, add the server URL to your AI tool, and authorize — live ad data in under 5 minutes.