Triple Whale MCP: the official server, and net profit next to ad spend

Yes. Triple Whale MCP is a read-only server Triple Whale runs at https://mcp.triplewhale.com/v1/mcp, documented in its help center under Triple Whale MCP, with setup guides for Claude and for ChatGPT written in July and August 2026. It lets external AI tools, Claude, ChatGPT, Cursor, n8n, Gumloop and others that support MCP servers, query and analyze your real-time Triple Whale data: performance, attribution, spend, return on ad spend, revenue and the other metrics Triple Whale computes.

Sign-in is OAuth2 scoped to moby:read, the method Triple Whale recommends wherever the tool supports it: add the server URL, sign in, select the business, allow read-only access. No API key or client secret is needed, and the access can be revoked at any time. For config-based or desktop tools without OAuth, such as Cursor, an MCP: Read API key generated on the Triple Whale APIs page is sent as an x-api-key header through a mcp-remote snippet.

Triple Whale MCP is included with the Foundation, Automate and Enterprise plans; a legacy plan that used the earlier connection may keep its access, and otherwise goes through the account team. The help center is careful to separate it from Moby, the AI teammate inside Triple Whale, and from API access: MCP access is not native API, headless-platform or custom integration access, which are separate capabilities with their own agreements. The developer portal, for its part, is about OAuth apps on the Triple Whale API.

The help center publishes no tool list for the server; it publishes the questions. The examples are a buyer's questions: what was my return on ad spend last week by channel, compare my ad spend on Meta versus Google this month and tell me which performed better, which products drove the most new-customer revenue in the last 30 days, how is my average order value trending this quarter, summarize yesterday's performance with revenue, orders, spend and ROAS. Moby answers them from your live Triple Whale data, so Moby usage must be available on your plan.

It never writes. Triple Whale states that the connection connects and analyzes your data and does not mutate or act on anything in your business, ad accounts, campaigns, budgets or Triple Whale account. Recurring check-ins, reports or draft copy built in your AI tool from MCP data live in that tool: they are not published, not connected to your business and cannot take action. Action workflows, native automations, Actions on connected platforms, landing pages and creative are Moby's job inside Triple Whale, with Automations, Actions and Landing Pages on the Automate plan and above.

Two practical notes from the guides: if Claude or ChatGPT asks for your business ID, give it the store's domain and ask again; and your questions are sent to Triple Whale so Moby can answer them, with OpenAI processing the prompts as well when the client is ChatGPT.

• Reads: real-time performance, attribution, spend, ROAS, revenue, orders, products and AOV, through Moby
• Writes: nothing; Actions, automations and landing pages stay inside Triple Whale
• Needs: an eligible plan, Moby usage, and a client that supports connectors or custom Apps

Adrails connects Triple Whale as a data source, not as a tool server, and asks it for one thing: the summary page of your store. In Triple Whale, create an API key and tick Summary Page: Read and nothing else; in Adrails, open Integrations, choose Triple Whale and paste the key. The key only reads, and disconnecting removes it.

One condition first: Triple Whale reports on a store, and Adrails needs exactly one connected Shopify store to point it at. With no store, or more than one, the connection is refused. From then on Adrails reads blended spend, ROAS, CPA, net profit and refunds every 6 hours, and reads no customer journeys. The figures appear in Analysis as the profit data of the chart, net profit, CPA, Meta ROAS and refunds, and as a Triple Whale row under Each integration, the same period as each source measured it, shown next to Meta, Google Ads and Shopify rather than merged into them.

An AI client connected to https://adrails.ai/api/mcp gets these series through metrics_daily, which returns the daily series the integrations measure, net profit, CPA and ROAS on the analytics side, and lists everything measured over the period when called without metrics, which is how to discover the names. integrations_status says what is connected and how fresh each data stream is; an owner or admin can queue a refresh with integrations_sync or rotate the key through a setup link from integrations_update.

Triple Whale's blended view and Meta's attributed view disagree by design, and the useful question is by how much and since when. Adrails holds both for the same dates. performance_period returns one ad account's spend, results, revenue, ROAS and CPA for an exact date range at the account, campaign, ad set or ad level, with a second range for the change, up to 400 days. metrics_daily returns the net profit, CPA and ROAS Triple Whale measured for the same days. Neither figure overwrites the other; the Analysis page keeps them as separate rows, and so do the tool results.

business_review checks the reconciliation once it is written: a bounded business explanation or calculation, unit economics, revenue reconciliation or budget pacing, from explicit inputs, where every number must quote its source word for word from the person's own statement or a tool result in the connection's history. metric_compare does the arithmetic of one observed value against a target. Ask whether net profit followed the spend increase on Meta this month: the answer names the Meta spend by campaign, the Triple Whale profit series, their periods, and stops where the data does.

The same figures feed the agent and the automations in Adrails: the agent gives profit figures when you ask for them, and automations run reports, alerts and pauses on a schedule or a threshold. An advertising change that follows, a pause or a budget move, is a proposal you apply in Adrails, capped at 30% per budget change, with an undo while the platform allows.

• Spend, ROAS and CPA by campaign for exact dates: performance_period
• Blended spend, ROAS, CPA, net profit and refunds by day: metrics_daily
• A reconciliation with every number sourced: business_review

Connect Triple Whale's server when the question is about Triple Whale's own numbers: attribution by model, new-customer revenue by product, the channel split as Triple Whale computes it, anything Moby already answers inside the platform. It is read-only, signs in with your Triple Whale account and business, and answers in the vocabulary of your dashboards.

Connect Adrails when the question has an ad account in it: which campaign's spend moved while net profit did not, whether the blended CPA and the Meta CPA drifted apart since a budget change, what to pause, and what the weekly review should say. Adrails reads Triple Whale's summary page next to Meta Ads and, once your workspace can connect it, Google Ads, with the Shopify orders the store recorded, and every advertising change is a proposal you apply in Adrails.

Both at once works in Claude or ChatGPT because the two servers do not overlap: one is the whole of Triple Whale, read through Moby; the other is five summary figures set beside the ad accounts and the store, with the tools to act on what they show. Each server signs in on its own account, Triple Whale's with moby:read, Adrails with a Summary Page: Read key that it holds for you.

For Triple Whale's server in Claude, open Settings, then Connectors under Customize, choose Add custom connector, name it Triple Whale, enter https://mcp.triplewhale.com/v1/mcp, continue with the detected OAuth client, connect, select the business and allow. In ChatGPT, an Admin or Owner on Business, or a user granted Developer Mode on Enterprise or Edu, opens Apps, creates an App with the same server URL and OAuth, then signs in and selects the business. Cursor and other config-based tools take the mcp-remote snippet with an MCP: Read API key.

For Adrails, add https://adrails.ai/api/mcp as a custom connector in Claude under Customize, then Connectors; run claude mcp add --transport http adrails https://adrails.ai/api/mcp in Claude Code; in ChatGPT, turn on Developer mode and add a custom MCP server with OAuth as its authentication. Sign in to Adrails, choose the workspace and the permissions, and allow access. Read access cannot be left out, and each connection works in one workspace.

Then connect Shopify in Adrails if it is not already, and Triple Whale after it, under Integrations, with a key limited to Summary Page: Read, or ask the assistant for a setup link with integrations_connect, valid for 10 minutes and opened by an owner or admin. Ask which integrations are connected: integrations_status lists Triple Whale with the freshness of its data, and metrics_daily without metrics returns the names of everything it measured.

• Triple Whale's server: https://mcp.triplewhale.com/v1/mcp, OAuth2 scoped to moby:read
• Adrails: https://adrails.ai/api/mcp as a custom connector, OAuth sign-in
• Triple Whale in Adrails: one Shopify store first, then a Summary Page: Read key