Claude or ChatGPT with an ads MCP server: the DIY route and its alternatives

Many B2B teams ask whether they need a tool at all. Claude or ChatGPT can connect to ad platforms through MCP, the Model Context Protocol, and answer questions about your ads in a chat. This is the DIY route. Teams look for an alternative when LinkedIn carries budget, because LinkedIn has no official MCP server. They also look when a chat forgets its read of each ad, or when nobody has time to keep the setup running.

We checked the facts on Meta's and LinkedIn's developer pages, on the public docs of the MCP servers and on the public pages of each tool below, on 22 October 2026. Each platform fact has a link. Where we could not find a point, the page says "Not found on their public pages". For the full list of tools, see the best creative analytics tools for LinkedIn and Meta.

The best alternatives to the DIY route for B2B

The tools after Nimblet are in no ranked order.

1. Nimblet. Nimblet is the finished version of the parts you would build, for B2B teams on LinkedIn and Meta. You connect LinkedIn with your normal login, with no developer app. It stores a read of each ad against your company, buyer and documents, so every chat starts from the same read of each ad. It reads thought leader and document ads, takes competitor ads from the LinkedIn and Meta Ad Libraries, and writes the next brief and drafts. It also runs inside Claude and ChatGPT through its own MCP server.
2. Motion. Motion is a creative analytics tool for brands that spend a lot on Meta. Its pricing page lists "full data & MCP access" on the Starter plan, from $999 a month. Its LinkedIn support is a beta for image and video ads. It fits if you buy mostly on Meta. See Motion alternatives.
3. Foreplay. Foreplay is a swipe file, ad search and brief tool. Its MCP server searches its ad library, your swipe file and your competitor watchlist, on the Workflow and Agency plans. Its reports on your own ads connect Meta and TikTok accounts. It fits if you mainly need to save and search other brands' ads. See Foreplay alternatives.
4. DatAds. DatAds is a creative strategy tool for DTC brands. Its pricing page lists "API / MCP" on the Regular and Custom plans. It connects Meta, TikTok and Google for your own accounts, and we found no LinkedIn account connection. See DatAds alternatives.
5. Segwise. Segwise covers more than 15 ad networks and generates finished ads. Its MCP server is read-only, has three tools and is turned on by your account manager. It fits if you buy on many networks. See Segwise alternatives.

The sections below show what the DIY route gives you, and how Nimblet differs.

What does Meta's ads MCP server give Claude or ChatGPT today?

Meta hosts an official ads MCP server that lets an assistant read your results, list your creatives, search the Meta Ad Library, and create and edit campaigns. Meta's overview says it runs at mcp.facebook.com/ads and "works with any MCP-compatible AI agent". The setup guide shows Claude Code and ChatGPT, with sign-in through Facebook Login for Business.

The tools fall into these groups:

• Reporting. Campaigns, ad sets and ads with spend, impressions, CTR, CPC, CPM and conversions, plus trends, anomalies and industry benchmarks, by the reporting tools page.
• Creatives. Tools that list the creatives, images and videos in an ad account and render an ad preview, by the ad creation page.
• Ad Library. ads_library_search searches "the public Meta Ad Library for competitive research and creative inspiration".
• Writes. Create campaigns, ad sets, ads and creatives. New items start paused, and your assistant asks before it turns one on.

This covers the needs of a Meta-only team. A price for the server: Not found on their public pages.

What can Claude or ChatGPT get from LinkedIn ads today?

LinkedIn has no official MCP server, so a DIY setup on LinkedIn uses a community server and your own API access. We found no MCP server in LinkedIn's marketing developer docs.

To use a community server, you need a LinkedIn developer app with access to the Advertising API. LinkedIn's access tiers page says this access "is reserved for applications building supported use cases". The Development tier lets you read data for the ad accounts you administer. LinkedIn "reserves the right to review applications". By LinkedIn's token page, access tokens last 60 days, and refresh tokens are for approved partners.

One example is an open-source community server. By its public README, read on 22 October 2026, it needs your own app with the Marketing Developer Platform, ships read-only tools, and reports results with breakdowns such as seniority. The README does not say that it returns ad images, videos or the post text of thought leader ads.

What does your team build around the connector?

The connector gives your assistant data. Your team builds the judgment around it. These are the parts Nimblet ships and a DIY setup has to make and keep up:

• A stored read of each ad. The assistant can look at an image in one chat. To compare many ads, it needs the same read of every ad, kept between chats: message, hook, offer, proof, audience, and for video the scenes and the spoken words.
• Your company and buyer in every answer. You paste your brand guide, product facts and sales notes into the chat or a project, and keep them current.
• A way to judge angles. Without a frame, the assistant names whatever pattern it sees. You write the frame, for example a set of persuasion mechanisms that work on B2B buyers.
• Competitors on both platforms. Meta's server searches the Meta Ad Library. LinkedIn competitor ads: Not found in the community server docs we read.
• Briefs and drafts in your brand. You write the prompts, the brand rules and the steps, and you keep the briefs and your team's feedback somewhere the next chat can find.
• Upkeep. Tokens expire, and API versions change. LinkedIn's tiers page, for example, says Marketing version 202511 sunsets on 16 November 2026.

When is DIY the right call?

DIY fits when you run Meta only and have someone who likes to build. Meta's server covers reports, creatives, the Ad Library and campaign changes.

DIY also fits when:

• You want the assistant to create or edit campaigns. Nimblet never does that.
• You ask one-off questions about a few ads, and a stored read adds little.
• You want to join your ad data with other sources in your own way.
• Your team has time to write prompts and keep the setup running.

How is Nimblet different from a DIY setup?

Nimblet is the finished version of the parts above, for B2B teams on LinkedIn and Meta. You ask Nib in the Nimblet app, or you ask Claude or ChatGPT through the Nimblet MCP server. The answer comes from a read that already exists.

LinkedIn without your own developer app

You connect LinkedIn with your normal login. Nimblet syncs your ads with images, video, copy and daily results. It types thought leader ads by their author and reads the person's post as the creative. When LinkedIn blocks a post, Nimblet reads the public post where it can. It reads the first page of each document ad and syncs document completions and download clicks. It shows who saw each ad, by seniority, job function, industry, company size and country. More on LinkedIn ads in Nimblet.

A stored read against your company

Nimblet reads your ads with spend in the background: images, videos scene by scene with a transcript, carousels, document ads and thought leader posts. It reads each one against the company, product, buyer and market you set, and against your team's documents and notes. Nib and every brief also draw on a built-in library of persuasion mechanisms for B2B buyers, such as loss aversion, risk reversal and social proof. The read is the same in every chat, for every person on the team.

Competitors on LinkedIn and Meta

Nimblet takes your competitors' ads from the LinkedIn Ad Library and the Meta Ad Library and reads them like your own ads. It writes a positioning summary per competitor and a weekly market news update. Ad libraries show no spend or results, and the Meta Ad Library shows ads that ran in the EU and EEA. See competitor ads.

The next brief and drafts, with a playbook

A brief starts from your best ads, your brand, your documents, your competitors' ads and the offer on your landing page. Nimblet drafts single images, carousels and video storyboards in your brand. Playbooks for LinkedIn and Meta batches guide Nib or your own assistant step by step, and the drafts your team reviewed shape the next brief. Anything that costs money or changes data waits until you confirm. See the AI ad generator.

One organization per client

Agencies get one organization per client. Ad accounts, brand and buyers stay apart, and your assistant works in one client at a time. See Nimblet for agencies.

How do DIY and Nimblet compare?

The table shows only the points that matter for a B2B team on LinkedIn and Meta.

What does each option cost?

A DIY setup costs your assistant plan and your team's time. Anthropic's connector guide says custom connectors work on Claude Free, Pro, Max, Team and Enterprise, with one custom connector on Free. On Team, an owner adds them. For ChatGPT, check OpenAI's help center for your plan. A price for Meta's server and a fee for LinkedIn's Advertising API: Not found on their public pages. The build effort is the list in the section above: the app, the access request, the prompts, the storage and the upkeep.

Nimblet has a Free plan that syncs and shows your LinkedIn and Meta ads. It has no AI analysis, no MCP and no competitor tracking. Basic reads over MCP. Pro and higher plans also write briefs and drafts, after you confirm. Nimblet shows paid prices on request. In ChatGPT, Nimblet needs a Business, Enterprise or Edu workspace with developer mode. See pricing.

Can you use Nimblet and Meta's server together?

Yes. Claude and ChatGPT can use several MCP servers in one chat. Nimblet gives the read, the competitors and the brief. Meta's server can then build ads in your account, paused, for you to check. When numbers differ between the two, check the conversion and the attribution window that each one uses.

Who should build it yourself?

Build it yourself in these cases:

• You run Meta only.
• You want your assistant to create and edit campaigns. Nimblet never does that.
• You have a technical person and time to keep the setup running.

Pick Nimblet if LinkedIn carries a real part of your budget, you need the same read of every ad in every chat, you want the next brief and drafts in your brand, or you run ads for several clients.