ResourcesDeep dives

LinkedIn Ads + Claude: the MCP that can actually manage campaigns

Daniel Popa, founder of AdPlug
Daniel PopaFounder, AdPlug8 min read

Updated July 14, 2026

Minimal flat illustration on warm cream: the coral Claude starburst character shaking hands with a slate briefcase character wearing a teal necktie, connected by a thin cable

If you run B2B paid media, you have probably noticed a gap: every AI-for-ads tool supports Google, most support Meta, and almost nothing touches LinkedIn, the platform where your actual pipeline comes from. This article covers what a LinkedIn Ads MCP is, why LinkedIn is the hardest ads platform to wire into an AI assistant, and what becomes possible when Claude can not only read your campaigns but create and edit them. I built the connector, so consider the bias declared; the workflows below all ran on my own ad account first.

Why LinkedIn is missing from AI ads tooling

Google Ads has an open, well-documented API and even an official open-source MCP server. Meta ships its own MCP. LinkedIn is different: access to the Marketing API goes through a developer-program review, write permissions require a separate tier, and the API itself has quirks that punish casual integrations. The result is that most tools stop at Google plus Meta and call it a day, which leaves B2B teams (the people watching cost per lead the hardest) without AI leverage on their most important channel.

AdPlug went through that review and holds Standard-tier access to the LinkedIn Marketing Developer Program, which is why the LinkedIn Ads MCP can offer both reads and writes on any account you manage in Campaign Manager, without you filing an API application or running a server.

What Claude can read

Once connected, the reporting surface covers the questions B2B teams actually ask:

  • Campaign, creative, and daily performance, plus conversion and lead-gen form reports
  • Demographics reports: cost per lead split by job seniority, function, industry, or company size, the report LinkedIn buries deepest in Campaign Manager
  • Audience counts and targeting lookups before you commit budget to a segment
  • The public LinkedIn Ad Library, which makes competitor research a prompt instead of an afternoon
Prompt
What's my cost per lead by job seniority on the Q2 demand-gen campaign group, last 90 days? Then compare it against Google Ads paid search for the same offer.

That second sentence is the quiet superpower: with Google Ads connected through the same setup, cross-channel questions get answered in one breath, no spreadsheet stitching.

What Claude can write (previewed)

Reading is table stakes. The part nobody else offers is the write surface: campaign groups, campaigns, saved and targeting audiences, lead-gen forms, ad creation across formats (single image, video, document, and more, including media upload), plus conversion events sent server-side in batches up to 5,000. Every mutation runs through a preview step: Claude shows the exact API operation, you confirm, then it executes. Write access is off by default and the free tier is read-only entirely.

Prompt
Create a campaign inside the Q3 Demand Gen campaign group: sponsored content, daily budget $80, targeting marketing directors at SaaS companies in the UK, excluding our customer list. Preview everything first; I will confirm before you execute.

A real workflow, end to end

A compressed version of a session I run on my own account most weeks:

Weekly LinkedIn Ads workflow with Claude, step by step
StepPrompt (abbreviated)What happens
1. PulseLast 7 days by campaign, flag CPL moversReport with movers ranked, 30 seconds
2. DiagnoseWhy did CPL jump on the webinar campaign?Demographics + creative breakdown, frequency check
3. CreativeWhich creatives fatigued? Draft 3 replacementsRanked fatigue list, copy drafts in brand voice
4. AudienceSize a lookalike-ish saved audience for the new segmentAudience count before any budget moves
5. ShipCreate the replacement ad, previewedPreview shown, confirmed, live
6. LogSummarise what changed for the client docSlack-ready changelog with numbers

Steps 1 through 4 are read-only. Step 5 is the one write, and it arrives as a preview I confirm. The whole session runs 20 to 30 minutes, most of which used to be Campaign Manager tab-hopping.

Setting it up

Five minutes: sign up, connect LinkedIn with OAuth (use the profile that has Campaign Manager access), paste the MCP server URL into Claude or ChatGPT, authorize, and ask it to list your ad accounts. The LinkedIn Ads connection guide walks every step with the exact URL, and the safety checklist explains the trust model (OAuth, encrypted tokens, audit log, instant revocation) if you need to clear it with a client or a security-minded boss first.

If you have been waiting for AI tooling to take LinkedIn as seriously as your pipeline does, this is that. Run the CPL-by-seniority prompt on your own account and see what it finds.

Questions, answered honestly.

What is a LinkedIn Ads MCP?

A hosted server that speaks the Model Context Protocol and exposes LinkedIn Marketing API capabilities as tools an AI assistant can call: performance reports, audience lookups, Ad Library search, and previewed campaign management. You connect it to Claude or ChatGPT by pasting one URL.

Why do so few tools support LinkedIn Ads?

LinkedIn gates its Marketing API behind a developer-program review, and write access requires a higher tier than read access. Most AI-for-ads tools stop at Google and Meta. AdPlug holds Standard-tier LinkedIn Marketing Developer Program access, which is what makes reads and writes possible for any account you manage.

Can Claude actually create LinkedIn campaigns and ads?

Yes: campaign groups, campaigns, audiences, lead-gen forms, and ads across formats including single image, video, and document, with media upload handled for you. Every creation or edit runs through a preview step you confirm before it executes.

Does it work for agencies managing multiple LinkedIn ad accounts?

Yes. One OAuth connection exposes every ad account your LinkedIn profile can manage in Campaign Manager, and prompts can scope to a specific account by name or ID. Every tool call lands in the audit log with account and timestamp.

Is my LinkedIn data used to train AI models?

No. Your campaign data flows from LinkedIn through AdPlug to your assistant to answer your questions, and is not retained for training or shared with model providers for training. Tokens are stored encrypted and you can revoke access at any time.

Try this on your account.

Connect Google Ads and LinkedIn Ads to Claude or ChatGPT in two clicks. Free tier, read-only by default, no card required.