Best MCP Servers for Google Ads and Paid Media in 2026

The best MCP servers for Google Ads and paid media in 2026 are the official ones from the ad platforms: Google Ads, Meta, Microsoft Advertising, Amazon Ads and TikTok. For reporting across many channels at once, third-party connectors such as Pipeboard, Supermetrics and Windsor.ai fill the gaps, including LinkedIn Ads, where we found no official server.

MCP (Model Context Protocol) is the open standard that lets an AI app such as Claude or ChatGPT use outside tools. Connected to an ad account, it lets you ask "Which campaigns spent over budget last week?" or, on servers that allow writes, "Pause every ad set with a cost per lead above $80."

We checked every entry against the vendor's own docs, GitHub page, announcement or pricing page on 2 November 2026. We did not test these products. Ad accounts spend real money, so read the write controls for each entry before you connect one.

How we chose

We used the MCP Guide's standard listicle criteria:

• Live and public. An MCP server documented or announced by the vendor itself, opened on 2 November 2026.
• Clear pricing. Free, included in a published plan, or a clear way to get pricing. Prices come only from official pricing pages.
• Security basics. The vendor explains sign-in (OAuth, tokens or service accounts) or permissions.

We split the list into platform-owned servers (entries 1 to 5) and third-party cross-channel connectors (entries 6 to 8), plus Google Analytics for conversion checks. We left out the many single-developer community servers for Microsoft, LinkedIn, TikTok and Pinterest ads because we could not confirm their maintenance or API approval. The order is not a ranking. For SEO and broader marketing tools, see the best MCP servers for marketing and the MCP Guide's best-of hub.

Paid media MCP servers compared

1. Google Ads MCP server: Google's official reporting server

What it is. Google's Google Ads MCP docs say "This implementation is strictly read-only" and that it "cannot modify bids, pause campaigns, or create new assets." It runs locally with pipx or on Google Cloud Run over Streamable HTTP. The core tools are search (GAQL queries), get_resource_metadata and list_accessible_customers.

A change to note. The GitHub README now lists four write tools: create_asset, upload_video_asset, manage_labels and manage_label_associations. Each uses a "Propose & Confirm" workflow: a dry run by default, then a confirm call with a checksum. The bundled config enables all tools, and the README says to set labels: false and assets: false for a strictly read-only server.

Best for: Search and Performance Max advertisers who want plain-English reporting.

Pricing: Free and open source (Apache 2.0). The docs say developer tokens "were sunset on September 9, 2026, and are no longer required." Normal Google Ads API limits apply.

Pros:

• Still cannot change bids, budgets or campaign status.
• Any write needs a two-step confirm with a checksum.
• Self-hosted, so your credentials stay in your own Google Cloud project.

Cons:

• Docs and README now disagree on write access; turn off the asset and label tools if you want read-only.
• Needs Google Cloud setup and GAQL knowledge to get the most from it.

Checked against: Google Ads MCP docs and README on 2 November 2026. Full profile: Google Ads MCP server.

2. Meta Ads MCP server: manage Facebook and Instagram ads

What it is. Meta hosts its server at https://mcp.facebook.com/ads. The Meta overview says it can "pull detailed reporting" and "Create and edit ads, ad sets, and campaigns." Tools fall into seven groups, including reporting, ad management, catalogs, signals and datasets, A/B and lift studies, and activity logs.

Best for: Advertisers and agencies that run most of their spend on Facebook and Instagram.

Pricing: Meta's docs list no MCP fee. Normal ad spend applies.

Pros:

• Official, hosted server with OAuth through Facebook Login for Business.
• Permission-based: writes need ads_management on top of ads_mcp_management.
• Activity logs and lift studies, not just spend reports.

Cons:

• It can create and edit live campaigns, so a mistaken prompt can spend money.
• Admin-role system user tokens "can't access the ads MCP server"; only Employee-role ones work.

Checked against: Meta get-started guide on 2 November 2026. Full profile: Meta Ads MCP server.

3. Microsoft Advertising MCP server: Bing and Microsoft ads data

What it is. Microsoft's MCP server page says that at launch you can explore account data (campaigns, ad groups, ads, keywords), view performance metrics, filter by date range and audit campaign setup. Microsoft points to a setup guide on Microsoft Learn for configuration.

Best for: Advertisers who run Microsoft Advertising alongside Google Ads.

Pricing: Free to connect. Microsoft notes you "may need a paid subscription for the AI assistants you use."

Pros:

• Official server from Microsoft Advertising.
• Supports Microsoft 365 Copilot via Copilot Studio, Claude, ChatGPT and "any MCP-compatible AI agent."
• Launch features focus on reading and auditing, which is a safe start.

Cons:

• The product page does not state the endpoint, sign-in method or whether writes are supported.
• We could not open the Microsoft Learn setup guide, so check it before you rely on details.

Checked against: Microsoft Advertising MCP page on 2 November 2026.

4. Amazon Ads MCP server: sponsored ads and DSP for partners

What it is. Amazon's open beta announcement, dated 2 February 2026, says the server is "available globally in open beta to Amazon Ads partners with active API credentials." It supports creating, updating and deleting campaigns, reporting, account settings and billing data. It includes multi-step tools, such as one that builds a full Sponsored Products campaign from one prompt.

Best for: Amazon sellers' agencies and tools that already use the Amazon Ads API.

Pricing: No MCP fee is mentioned in Amazon's announcement.

Pros:

• Official server covering sponsored ads, DSP and reporting.
• Pre-built workflow tools reduce the number of steps an AI has to chain.
• Global availability.

Cons:

• Open beta, limited to partners with active API credentials.
• It can delete campaigns and read billing data, so scope access carefully.

Checked against: Amazon Ads open beta announcement on 2 November 2026.

5. TikTok Ads MCP server: announced for agent-run campaigns

What it is. TikTok's TikTok World 2026 newsroom post, dated 13 May 2026, introduces the TikTok Ads MCP Server, "a new interface that allows developers and advertisers to build AI agents and tools." It launched with TikTok Ads Skills, building blocks for "campaign creation, performance insights, creative analysis, audience discovery, and budget optimization."

Best for: Brands and agencies with meaningful TikTok spend that want to build their own agent workflows.

Pricing: Not published.

Pros:

• Official server from TikTok.
• Paired with ready-made Skills for common ad jobs.
• Aimed at both developers and advertisers.

Cons:

• We could not open public developer docs for the endpoint, sign-in or tool list.
• The newsroom post does not state availability by region or account type.

Checked against: TikTok newsroom post on 2 November 2026.

6. Pipeboard: one connector for many ad platforms

What it is. Pipeboard is an independent hosted service that connects Meta, Google, TikTok, LinkedIn, Microsoft, Pinterest, Snap and Reddit ads to MCP clients. Its open-source Meta Ads MCP server has about 1.3k GitHub stars and includes tools to create and update campaigns, ad sets, ads and creatives. It calls itself "an independent open-source project that uses Meta's public APIs."

Best for: Agencies managing several ad platforms, including LinkedIn, which has no official server.

Pricing: The Pipeboard pricing page lists Free (30 AI executions a week, 2 ad accounts), Pro at $29.90 a month, Premium at $99 and Agency at $199. Extra ad accounts cost $10 a month each.

Pros:

• Covers platforms the vendors do not, such as LinkedIn and Pinterest.
• Permission-scoped API tokens from the Pro plan up.
• Google Analytics and Search Console connections don't use an ad account slot.

Cons:

• A third party holds your ad platform credentials server-side.
• Bulk operations need Premium or higher.

Checked against: Pipeboard pricing and Meta Ads MCP README on 2 November 2026.

7. Supermetrics MCP: enterprise cross-channel reporting

What it is. Supermetrics' MCP docs describe a server for "AI applications and agents" that reaches Google Ads, Facebook, TikTok, Google Analytics, LinkedIn Ads and more. It answers performance questions and runs pacing checks and alerts. It can also "Create or modify campaigns," and new campaigns start paused.

Best for: Larger marketing teams that already pipe ad data through Supermetrics.

Pricing: The MCP product page says MCP is included in plans with Data API access. The Supermetrics pricing page shows Data API access only on Enterprise, which is priced by quote. A 14-day trial is offered.

Pros:

• Many data sources through one connection.
• New campaigns are created paused.
• Works with n8n and custom agents as well as chat apps.

Cons:

• Effectively an Enterprise-tier feature.
• Write access to campaigns means you need clear approval rules.

Checked against: Supermetrics MCP docs on 2 November 2026.

8. Windsor.ai MCP: low-cost blended reporting

What it is. Windsor.ai hosts its server at https://mcp.windsor.ai/. Its GitHub README says it reaches 350+ data sources, signs in with OAuth 2.0 and offers write tools (list_actions, execute_action) on connectors that support them, such as pausing ad campaigns.

Best for: Small teams and freelancers who want blended ad reporting on a budget.

Pricing: The Windsor.ai pricing page says every plan includes Windsor MCP. Basic is $23 a month ($19 billed yearly) for 3 data sources; Standard is $118 ($99 yearly). A forever-free plan allows 1 data source and 1 account.

Pros:

• Lowest paid entry price on this list.
• Open-source (MIT) repository with setup docs.
• The README tells clients to confirm with the user before write actions.

Cons:

• Plans cap data sources and accounts.
• Write actions depend on each connector, so test what is possible before granting it.

Checked against: Windsor.ai MCP README on 2 November 2026.

9. Google Analytics MCP server: check what happens after the click

What it is. Google's Analytics MCP docs describe a local server that runs GA4 standard, realtime, funnel and conversion reports. Google says it "can't create, modify, or delete your Google Analytics configuration or settings." The README adds one opt-in write tool for annotations behind an --allow-write flag.

Best for: Matching ad spend to on-site conversions and landing page performance.

Pricing: Free and open source (Apache 2.0).

Pros:

• Read-only scope by default.
• Lists your Google Ads links, so you can check the ad-to-GA4 setup.
• Funnel reports show where paid visitors drop off.

Cons:

• Local install with Google Cloud credentials.
• Conversion numbers may differ from each ad platform's own attribution.

Checked against: Google Analytics MCP docs on 2 November 2026. Full profile: Google Analytics MCP server.

Guardrails for ad accounts

Ad accounts are the riskiest place to give an AI write access, because mistakes cost money right away.

1. Read first. Start with reporting only. Turn off Google's asset and label tools, and grant Meta only ads_read at first.
2. Approve every spend change. Keep budget, bid and status changes behind a person. Our tool permission planner helps you decide which tools need approval.
3. Prefer paused drafts. Supermetrics creates campaigns paused. Ask other tools to do the same and review before launch.
4. Check the logs. Meta exposes activity logs; review them after the first week.
5. Reconcile numbers. Ask the AI to cite the date range and attribution window it used.

For a weekly reporting routine, see our weekly marketing report playbook.

Next step

If you need a custom MCP server for an ad platform without one, or tighter controls than a vendor server offers, our MCP development team builds and hardens servers for business systems.

Frequently asked questions