How to Connect LinkedIn Ads to Claude
VigilDog Team · October 8, 2026 · 6 min read
LinkedIn Campaign Manager is powerful, but every routine task, reading last week's numbers, spinning up a sponsored-content test, pausing a bleeding campaign, buries you in tabs. Connecting LinkedIn Ads to Claude lets you drive the same account in plain language, with each change proposed as a dry run you approve before it ever touches spend. Here's how the connection works and how to set it up without handing an AI the keys to your budget.
What connecting LinkedIn Ads to Claude actually means
Claude can't reach LinkedIn on its own. The bridge is an MCP (Model Context Protocol) server, a small service that holds your LinkedIn authorization and exposes the ad account as a set of typed tools Claude is allowed to call: list campaigns, pull analytics, draft a creative, adjust a budget. When you ask Claude to "show me last week's lead-gen cost per result," it doesn't guess; it calls the reporting tool and reads back real API data.
The important nuance: a well-built MCP separates reading from writing. Reads run freely. Anything that changes money or live campaigns is staged first as a preview, and nothing is submitted to LinkedIn until you explicitly confirm. That approval gate is the whole reason this is safe to use on a real account instead of a sandbox.
What you need before you start
The connection assumes a few things are already in place. None of them are exotic, but skipping one is the usual reason a first attempt stalls.
- A LinkedIn ad account where you hold an Account Manager or Campaign Manager role, Viewer access can read but not launch or edit.
- Access to Claude (a paid plan or the desktop app) or ChatGPT, both of which support connecting MCP servers.
- A VigilDog account to run the Ads MCP, which brokers the LinkedIn OAuth and enforces the approval gate.
- Your billing and payment method already configured inside Campaign Manager, the MCP can create and manage campaigns, but LinkedIn still requires valid billing on the account itself.
Connecting the account, step by step
The flow is deliberately short. In your VigilDog dashboard, open the Ads MCP and choose LinkedIn as the platform to connect. You'll be redirected to LinkedIn's own OAuth screen, this is the only place credentials are entered, and they go to LinkedIn, not to Claude. Approve the requested advertising scopes and pick the ad account you want to manage. Control returns to VigilDog, which stores an encrypted, revocable token; your password is never seen or held.
Next, add the MCP server to your assistant. In Claude, open Settings, go to Connectors, and add the server URL VigilDog gives you. In ChatGPT, add it the same way under custom connectors. Once it appears as connected, start a new conversation and ask Claude to list your LinkedIn campaigns. If names and statuses come back, the pipe is live. If you manage several accounts, name the client in your prompt so Claude targets the right one, the same approval-gated Ads MCP handles Google, Meta, and LinkedIn side by side.
Your first safe commands
Resist the urge to launch a campaign on minute one. Spend the first session reading, so you learn how Claude interprets your account and you build trust in the numbers it returns. Good opening prompts are diagnostic, not destructive.
Try: "List active LinkedIn campaigns with spend and CPL for the last 14 days," then "Which sponsored-content ads have the highest engagement but worst conversion rate?" These are read-only tool calls, no preview, no risk. When you're ready to change something, phrase it as an outcome ("pause any campaign over a $120 cost per lead") and let Claude assemble the edit for your review.
- Pull a plain-English performance summary across campaign groups.
- Ask for underperformers by objective (lead gen, brand awareness, website visits).
- Draft, not publish, a new single-image ad or a variant of an existing one.
- Model a budget shift between two campaigns before committing it.
How the approval gate keeps spend safe
Every write operation follows the same two-step ritual. Claude first calls a preview: it returns exactly what will change, the campaign, the old value, the new value, the daily budget impact, and hands you a single-use token. Nothing has hit LinkedIn yet. Only when you confirm does the MCP submit the change, and the token can't be replayed or tampered with to slip a different edit through.
This matters more on LinkedIn than most platforms, because LinkedIn's minimums are high, daily budgets and bids are real money fast. The dry-run pattern means a misheard "increase budget to 500" surfaces as a clear diff you can reject, not an overnight surprise. You get the speed of natural language with the discipline of a change-review step, which is the trade every serious advertiser actually wants.
Where this fits in your workflow
Connecting LinkedIn Ads to Claude doesn't replace Campaign Manager for deep audience building or creative review, it replaces the tedious 80%: the daily reads, the routine pauses, the budget nudges, the "what changed since Friday" questions. Agencies running several accounts feel it most, because the same conversation can hop from LinkedIn to Google to Meta without re-learning three interfaces. If you already run Google Ads from Claude, adding LinkedIn is the same muscle memory.
Set it up once, keep the read-only habit for a week, and let the approval gate carry the risk. From there, most of your reporting and routine optimization moves into a chat window, auditable, reversible, and fast.