Which third-party MCP servers connect Google, Meta and LinkedIn ads to AI assistants?

TofuAnswer by the Tofu team · checked

An orange braided cable splitting into four different connector heads against a blue background

We confirmed eight on the vendors' own sites. GoMarble, Pipeboard, Adspirer, Porter Metrics, Markifact, Windsor.ai and Supermetrics connect Google, Meta and LinkedIn ads; AdPlug connects Google and LinkedIn but not Meta. All eight run hosted servers, and all can make some changes once you allow it. Use the official servers first where they fit: Meta's own server for Meta, and Google's read-only server for Google Ads reports. A third party makes sense for LinkedIn, for changes to Google Ads, or for one connection across platforms.

As of October 2026.

How do the official and third-party ads MCP servers compare?

Prices are as each vendor publishes them, checked October 26, 2026. "No" under open source means we found no published server code.

ServerPlatformsRead or writeHostingPublished priceOpen source
Google Ads MCP server (official)Google Ads onlyRead onlyYou run it, locally or on Cloud Run; a hosted pilot is open to allowlisted participantsFreeYes, Apache 2.0
Meta ads MCP server (official)Meta onlyRead and write; new ads start pausedHosted by Meta at mcp.facebook.com/adsNo price listedNo
GoMarbleGoogle, Meta and LinkedIn (LinkedIn is reporting only)Read only on Copilot; changes with your approval on Operator and up, for Google and MetaHostedCopilot $139/mo, Operator $399/mo, Scale $799/mo; 7-day trialHosted server no; separate MIT-licensed, read-only Google Ads and Facebook Ads servers you run yourself
PipeboardGoogle, Meta, LinkedIn, plus TikTok, Microsoft, Pinterest, Snap and RedditRead and write; read-only and per-tool tokens on Pro and upHosted; its Meta server can also run locallyFree (2 ad account slots, 30 tool runs a week), Pro $29.90/mo, Premium $99/mo, Agency $199/moMeta server is source-available under the Business Source License 1.1
AdspirerGoogle, Meta, LinkedIn, plus TikTok, Amazon and ChatGPT adsRead and write; new campaigns start paused, and it can't delete existing campaignsHostedFree (15 tasks a month), Plus $49/mo, Pro $99/mo, Max $199/mo; less when billed yearlyNo
Porter MetricsGoogle, Meta, LinkedIn and 26 platforms in allRead; changes after you confirm each one, with new campaigns and creatives pausedHostedFree plan; reporting from $15 a month for one account ($12.50 billed yearly), less per account as you add more; changes use credits, 100 a month freeNo
MarkifactGoogle, Meta, LinkedIn, plus TikTok, Microsoft and moreRead and write; changes need your approvalHosted in the EUPro from $79/mo, Team from $249/mo; trial creditsNo
AdPlugGoogle and LinkedIn, plus Microsoft and Reddit (no Meta)Read only by default; writes are a per-platform switch, with a preview of each callHostedFree (one platform, view only), Pro $29/mo (200 edits), Max $89/moNo
Windsor.aiGoogle, Meta, LinkedIn and hundreds of other sourcesRead; changes on Google Ads, Meta Ads, LinkedIn Ads and other platforms that support them, after you confirm each, and you can turn writes offHostedFree (one data source), Basic $23/mo, Standard $118/mo; less when billed yearlyNo
SupermetricsGoogle, Meta, LinkedIn, plus TikTok, Microsoft and moreRead; its docs say it can also create or change campaigns, with new ones pausedHostedPlans with Data API access, which its pricing page lists under Enterprise (quote); 14-day trialNo

General tool platforms offer ad connectors too. Composio, which connects AI agents to more than 1,500 tools, lists Google Ads, Meta Ads and LinkedIn Ads toolkits, with a free Hobby plan (100,000 tool calls a month) and Pro at $29/mo. It isn't an ads product, so check what approval steps your own agent setup has before you allow writes.

When is an official server the right pick?

  • Google Ads reports. Use Google's Google Ads MCP server. It's free, open source and runs under your own Google Cloud project. Google's developer policies ban any hosted "MCP server, proxy endpoint, or any similar service that solely replicates, wraps, or re-exposes Google Ads programmatic capabilities," while they don't restrict "open-source tools where the end-users download the software and connect it to Google Ads using their own access credentials." If you use a hosted vendor for Google Ads, ask how it fits that policy.
  • Meta. Use Meta's ads MCP server if Meta is your main platform. It reads and builds campaigns, new ads start paused, and you can block actions per ad account in Meta Business Suite. A third party adds Meta to the same connection as your other platforms.
  • LinkedIn. LinkedIn's Marketing API change log, updated October 2026, doesn't mention an MCP server, so a third party is the way to connect today. See the LinkedIn Ads MCP server guide for the current status.

For a quick pick by situation, see whether you can connect Claude or ChatGPT to your ad accounts.

When is each third-party server the right pick?

GoMarble

Pick it if you want scheduled agents, creative analysis and competitor research on Meta and Google, with approvals and an audit trail. Changes need the Operator plan, and LinkedIn is reporting only. If you searched for GoMarble's Google Ads MCP server on GitHub, that's a separate open-source project: read only (accounts, GAQL queries, Keyword Planner ideas), MIT-licensed and run on your machine. Its setup still asks for a Google Ads developer token, which Google sunset on September 9, 2026.

Pipeboard

Pick it for a low entry price and eight ad platforms behind one sign-in. Tokens can be read only or limited to chosen tools, and the Agency plan scopes tokens per ad account. Pipeboard says new Google Ads campaigns start paused and LinkedIn creatives start as drafts. Its Meta server code on GitHub allows production use but not a competing hosted service, and converts to Apache 2.0 on January 1, 2029.

Adspirer

Pick it if you want an agent that plans and launches campaigns across Google, Meta, LinkedIn, TikTok, Amazon and ChatGPT ads, priced by task rather than by ad spend. Adspirer's platform pages say new campaigns are created paused and that it can't delete existing campaigns. Its servers are listed in Claude's connector directory, where one listing also names Microsoft Advertising; Adspirer's own site doesn't list it as a platform.

Porter Metrics

Pick it if you mostly report and sometimes act. You pay per connected account for reporting, and each completed change costs one credit. Porter says Claude creates LinkedIn campaign groups, campaigns and ads as drafts and changes budgets or targeting "each after you confirm."

Markifact

Pick it for approval-gated changes across 50+ marketing platforms, EU hosting and execution logs (7 days on Pro, 30 on Team). On Team you can connect Google Ads and Meta through your own apps, so client change history shows your app, not Markifact's.

AdPlug

Pick it for Google Ads and LinkedIn with writes off until you switch them on. AdPlug says every write shows the exact API call before it runs, and it logs every change. It doesn't connect Meta.

Windsor.ai

Pick it if you mainly want reporting across many sources at a low price. Windsor.ai says writes work on Google Ads, Meta Ads, LinkedIn Ads and other platforms that support them, each after you confirm, and you can switch them off for read-only access.

Supermetrics

Pick it if your team already pays for Supermetrics with Data API access. Its pages differ on who gets MCP: the product page offers MCP access "with any subscription," its FAQ says plans "that include Data API access," and its docs say a subscription "with access to the Supermetrics API." Its pricing page lists Data API access only on Enterprise, so confirm before you buy.

Which ads MCP servers are on GitHub?

  • googleads/google-ads-mcp: Google's official Google Ads server, Apache 2.0, read only.
  • gomarble-ai/google-ads-mcp-server and gomarble-ai/facebook-ads-mcp-server: GoMarble's MIT-licensed, read-only servers you run yourself. The Facebook one had security fixes in July 2026.
  • pipeboard-co/meta-ads-mcp: Pipeboard's Meta server with read and write tools, under the Business Source License 1.1. Its README recommends the hosted version, and a local install needs your own Meta developer app.

Meta's official server isn't on GitHub; Meta hosts it.

What should I check before giving a third party write access?

  • Scopes. On Meta, a token needs ads_mcp_management plus ads_read or ads_management, and ads_management gives the agent more tools, so grant ads_read for reporting. On LinkedIn, r_ads and r_ads_reporting read, while rw_ads can "manage and read" ad accounts. Google Ads has a single OAuth scope, https://www.googleapis.com/auth/adwords, so limit the Google account instead: a Read-only user can view campaigns and run reports but can't edit campaigns.
  • Accounts. Connect only the ad accounts the tool needs. For agencies, check that each teammate and each token can reach only their clients' accounts.
  • Approvals. Check what starts paused or as a draft, whether every write needs a confirm step, and who can approve.
  • Audit logs. Ask how long the vendor keeps a record of each change and who made it. Then check the platforms' own records too: change history in Google Ads and campaign history in Meta Ads Manager.
  • Google's policy. Ask any hosted vendor how its Google Ads access meets Google's August 2026 developer policies. Can an AI agent get your ad account banned? covers what Google and Meta enforce.
  • Revoking access. Know where to cut access: in the vendor's app and in each ad platform. GoMarble, for one, says access "can be revoked from GoMarble or the platform."
Sources (45)

Ask Tofu next

Pick a question and Tofu Ads Agent answers it in chat.

Related answers

More on