Kiin Intelligence - LinkedIn Ads MCP

LinkedIn Ads analysis via MCP with benchmarks from 1,000+ accounts

Servidor MCP hospedado

npx add-mcp 'https://mcp.kiin.co/mcp'

Instala no Claude Code, Codex, Cursor e outros

Documentação

What is a LinkedIn Ads MCP server?

It is a service that lets an AI assistant call named tools against your LinkedIn ad account, such as “get campaign metrics” or “pause campaign”, using the Model Context Protocol. You ask in plain English; the assistant decides which tools to call and answers from live data instead of guessing.

Model Context Protocol is an open standard for connecting AI assistants to outside systems. Anthropic published it in late 2024 and OpenAI, Microsoft, Google and xAI clients now support it. An MCP server is the thing on the other end of the connection. You never see the tool names. You ask a question and the assistant works out which calls to make, and it queries again when you ask a follow-up, which is why the second and third questions in a session are usually the valuable ones.

MCP is not “AI in your dashboard”

Most tools sold as AI reporting run a model over a fixed export and give you a summary. With MCP the assistant holds the conversation and pulls exactly the data it needs, when it needs it, across as many calls as the question takes.

What can Kiin Intelligence do?

Kiin Intelligence registers 95 MCP tools. This is what they cover. Writes need Analyst access on that ad account (paid, or during the 14-day trial), and every write is logged.

AreaWhat it coversExample tools
Account reportingAccount snapshot, reporting overview, time series, campaign metrics and setup, budgetsget_account_snapshot, get_reporting_overview, get_time_series, get_campaign_budgets
CreativesCreative-level performance, creative health and fatigue, top creativesget_creative_health_report, get_top_creatives
Format reportsThought leader ads by author and hook, conversation ads (opens, not clicks), lead gen forms (submissions, not opens), videoget_tla_report, get_conversation_ads_report, get_lead_gen_report, get_video_performance_report
Audience and wasteDemographics, demographic waste against your ICP, company engagement (which companies saw or engaged), reach and frequency, audience inventory, targeting searchget_demographic_waste, get_company_engagement, get_reach_report, search_targeting
Audits and recommendationsAccount hygiene and full audit, prioritised recommendations, bid recommendations, Kiin Brain hard rulesrun_audit, get_account_hygiene, get_recommendations, get_bid_recommendations, get_hard_rules
BenchmarksYour numbers against the Kiin Labs panel: 1,000+ advertiser accounts, $65M+ of LinkedIn spend, no cohort under 10 advertisersget_benchmarks, benchmark_me
Writes (preview-then-confirm)Create, duplicate and update campaigns (budgets, bids, schedule, status, targeting, Audience Network and Audience Expansion switches); pause and reactivate; create and update campaign groups and creatives; create saved audiences. New campaigns are drafts.create_campaign, duplicate_campaign, update_campaign, patch_campaign_targeting, pause_campaign, reactivate_campaign, create_campaign_group, update_campaign_group, create_creatives, update_creative, create_saved_audience
CRM and pipelineHubSpot, Salesforce, Pipedrive and Attio. Influenced pipeline and revenue: deals at companies your ads reached or engaged before the dealget_pipeline_report, get_deals, get_contacts
Memory and monitoringSaved ICP and account strategy; scheduled monitoring with Slack alertsget_icp, get_account_strategy, set_monitoring

Can a LinkedIn Ads MCP server change campaigns, or only read them?

Kiin reads and writes. It can create, duplicate and update campaigns, campaign groups, creatives and saved audiences, change budgets, bids, schedules and targeting, and pause or reactivate campaigns. Every write is preview-then-confirm: the first call returns a before/after diff and changes nothing.

Only a second, confirmed call applies the change. New campaigns are created as drafts, so you activate them in Campaign Manager. Writes need Analyst access on that ad account (paid, or during the 14-day trial), and every write is logged. Kiin's built-in guardrails keep the LinkedIn Audience Network and Audience Expansion off and avoid raising bids on fatigued campaigns.

That matters because many comparison pages still describe Kiin as read-only. It is not.

Why does LinkedIn Ads data need a metric layer?

Because LinkedIn's clicks field is not traffic. It counts likes, comments, shares, follows and profile clicks as well as visits to your landing page. At the median account only about 1 in 7 LinkedIn clicks reaches your website, so CTR and CPC built on clicks flatter every campaign.

The field for traffic is landingPageClicks. Kiin reports landing page CTR (landingPageClicks ÷ impressions) and cost per landing page click (spend ÷ landingPageClicks) as the primary traffic metrics for feed ads, and shows total CTR only as context. The same discipline applies across formats:

  • Lead gen forms: form opens are not leads. Cost per lead uses submissions (oneClickLeads) only.
  • Conversation ads: judged on opens and open rate, not clicks.
  • Video: view rate and completion rate, not clicks.

Fluent and wrong is worse than obviously wrong

If a model reads a mislabelled field, it builds a confident argument on top of it. Bad data plus fluent writing is more dangerous than bad data in a spreadsheet, because nobody re-checks the arithmetic on something that reads that cleanly.

What does the Kiin audit check?

run_audit runs the Kiin Brain across eight layers and returns a prioritised list of what to fix, ordered by what would move results rather than by what looks worst. Each layer produces a score and an action; the health score is a weighted average of all eight.

  1. Thought leader ads: presence, volume, author diversity, landing page CTR, engagement rate
  2. Conversation ads: open rate, click-to-open rate, cost per lead, send volume
  3. Lead gen forms: submit rate, cost per lead, form friction
  4. Single image and company ads: landing page CTR, CPM, cost per landing page click
  5. Video ads: view rate, completion rate, cost per view
  6. Full-funnel structure: coverage and budget across top, middle and bottom of funnel
  7. GTM and outbound alignment: ICP targeting quality, company targeting
  8. Content strategy: creative diversity, message variation, format mix

The rules behind it come from Kiin Labs research on the panel, such as the finding that maximum delivery cost 103% more per website click than a manual bid in the same account.

Which AI assistants can connect to LinkedIn Ads through Kiin?

Claude, ChatGPT, Copilot, Grok and any MCP-capable assistant can connect. Claude and ChatGPT are documented in Kiin's own setup screen. The other paths below come from each client's current official docs (checked 26 Oct 2026). Where a client cannot add a remote OAuth server, we say so.

ClientStatusHow to add Kiin
Claude (web and desktop)SupportedCustomize → Connectors → + → Add custom connector → paste https://mcp.kiin.co. Team and Enterprise: an owner adds it under Organization settings → Connectors first. The Free plan allows one custom connector.Guide
ChatGPTSupported, paid planSettings → Connectors → Add custom connector (newer: Plugins → + → Add custom MCP server) → paste the URL → sign in → enable Kiin in a new chat. Some plans need Developer mode or an admin to allow custom connectors.Guide
Claude CodeSupportedclaude mcp add --transport http kiin https://mcp.kiin.co/mcp, then /mcp to sign in.Guide
CursorSupportedAdd {"mcpServers":{"kiin":{"url":"https://mcp.kiin.co/mcp"}}} to ~/.cursor/mcp.json; Cursor runs the OAuth sign-in.Guide
Microsoft Copilot StudioSupportedAgent → Tools → Add a tool → New tool → Model Context Protocol → URL → OAuth 2.0, Dynamic discovery.Guide
GitHub Copilot in VS CodeSupportedCommand Palette → MCP: Add Server → HTTP → paste https://mcp.kiin.co/mcp.Guide
Microsoft 365 CopilotAdmin onlyAdmin center federated connector. Microsoft documents a pre-registered OAuth client, not dynamic registration, so ask us before you try it.Guide
Consumer Microsoft Copilot appNot possiblecopilot.microsoft.com only offers Microsoft's own connectors (OneDrive, Outlook, Google). No custom MCP option.Guide
GrokSupportedgrok.com/connectors → New Connector → Custom → paste the URL. Business and Enterprise: an admin provisions it first.Guide
Gemini CLISupportedgemini mcp add --transport http -s user kiin https://mcp.kiin.co/mcp, then /mcp auth kiin.Guide
Gemini appUS personal accountsSettings → Connected Apps → Custom apps → Add a custom app. Google limits this to 18+, US, personal (not work) accounts, on the web.Guide
PerplexitySupportedPlugins (or Customize → Connectors) → Add custom → Connector → Add MCP connector → Remote. Enterprise: an admin allows custom connectors.Guide
Any other MCP clientSupportedAny client that can add a remote MCP server with OAuth. Use https://mcp.kiin.co/mcp (streamable HTTP) or https://mcp.kiin.co/sse for SSE-only clients.Guide

What can you ask a LinkedIn Ads MCP server?

Anything you would otherwise answer with three exports and a pivot table. Start with a snapshot, then ask about waste, warm accounts and fixes. The pattern that works is one question and two follow-ups: the first answer says what happened, the follow-ups find out why.

  • “Give me an account snapshot for last month, with landing page clicks and cost per lead.”
  • “Run an audit and give me the three things to fix this week.”
  • “How much of last month's spend went to job titles outside our ICP?”
  • “Which companies engaged more than three times but never clicked through?”
  • “Compare our thought leader ads with our single image ads on cost per landing page click.”
  • “How does our CPM compare with similar advertisers?”
  • “Pause the two campaigns with the worst cost per lead. Show me the change first.”
  • “Which deals last quarter were at companies our LinkedIn ads reached before the deal?” (needs a CRM connection)

Does LinkedIn have its own official Ads MCP server?

Yes, but it is undocumented and gated. LinkedIn runs a first-party Marketing MCP endpoint at https://mcp.linkedin.com/marketing/mcp. It answers with an OAuth challenge and advertises ads read and write scopes, but LinkedIn has published no docs, announcement or tool list, and it needs an approved LinkedIn developer app.

Most guides still say LinkedIn has no hosted Ads MCP. The difference is between a public product and an endpoint that exists. These are the facts from our own probes (7 Oct 2026, repeated 26 Oct 2026), not from a blog.

PropertyWhat we observed
Endpointhttps://mcp.linkedin.com/marketing/mcp
Response to an MCP initialize requestHTTP 401 with www-authenticate: Bearer resource_metadata=…, the standard MCP OAuth challenge
Proof it is first-partymcp.linkedin.com is a CNAME to mix-cdn.linkedin.com; responses carry x-li-fabric: prod-lor1, a LinkedIn production header
Advertised scopesr_ads, r_ads_reporting, rw_ads, r_organization_social, w_organization_social: ads read, ads reporting and ads write
AuthOAuth with PKCE against https://www.linkedin.com/oauth/pkce
Access modelNo dynamic client registration, so a client needs a pre-registered LinkedIn app: in practice an approved Marketing Developer Platform application
Siblings/learning/mcp returns 401, /talent/mcp returns 403, /ads/mcp and /mcp return 404
DocumentationNone found as of 26 Oct 2026: no docs, announcement or changelog entry

We cannot list its tools: that needs a token from an approved app. The full write-up, with a curl command you can run yourself, is on does LinkedIn have an MCP server? The only public LinkedIn surface we found is a beta, read-only “LinkedIn Ads” app inside ChatGPT that shows ad set performance.

When LinkedIn's server is the right answer

If you already hold an approved Marketing Developer Platform app and want raw API access through an assistant, with no third party in the data path, LinkedIn's endpoint is the natural choice once it is opened up.

When it is not

A passthrough returns what the API returns. LinkedIn's clicks field counts reactions, comments, follows and profile visits alongside visits to your site; at the median account only about 1 in 7 clicks reaches your website. An assistant reading a passthrough will build CTR and CPC on clicks unless something tells it not to. It also has no comparison set, so it can tell you your CPM but not whether it is good. Kiin returns landing page metrics as primary and benchmarks them against 1,000+ advertiser accounts and $65M+ of spend.

Which LinkedIn Ads MCP servers exist, and how do they compare?

As of 26 Oct 2026 there are at least 20 hosted LinkedIn Ads MCP servers and AI connectors, plus open-source repositories. They differ most on four things: whether they write, whether they benchmark, whether they audit, and whether they join ads to CRM pipeline. Kiin built one of these and lists it first.

ToolBenchmarksPlaybooks / insightsWritesAuditsCRM pipelineSetupPricing as stated
Kiin Intelligence LinkedIn Ads only1,000+ accountsKiin BrainPreview-then-confirmYes4 CRMsRemote URL + LinkedIn sign-inFree on 1 account; $65/account/mo
ZenABM LinkedIn ABM platformNot stated15 skillsYesYesHubSpot, Salesforce, AttioRemote, OAuthFrom $59/mo; 37-day trial
AdPlug Multi-platformNot statedNot statedPreviewed editsNot statedNot statedHosted, LinkedIn OAuthFree (view only); $29, $89/mo
PaidSync Multi-platformNot statedNot statedDry run defaultNot statedNot statedRemote, OAuth or API keyFree 15 tasks; from $99/mo
Markifact Multi-platformNot statedNot statedHeld for approvalNot statedNot statedRemote, OAuth 2.1Free credits; $79/mo
Ryze AI Multi-platformNot statedNot statedApproval-gatedNot statedNot statedRemote, LinkedIn OAuth$89/mo
Adspirer Multi-platformNot statedNot statedCampaigns created pausedNot statedNot statedRemote, OAuth 2.1Free 15 tasks; paid from $40/mo
Opus Growth Multi-platformNot statedNot statedPreview then approveNot statedNot statedRemote, LinkedIn OAuthFree trial; $29, $49, $99/mo
AdKit Multi-platformNot statedNot statedDrafts until approvedNot statedNot statedNot statedFrom $29/mo; 7-day trial
Insightful Pipe Multi-platformNot statedAuditor skillNew campaigns pausedSeparate skillNot statedHosted; MIT repoFrom $29.99/mo
Windsor.ai Multi-sourceNot statedNot statedPause, budgets, targetingNot statedNot statedHostedFree for 1 source; from $19/mo
Synter Multi-platformNot statedNot statedPause/enableNot statedCRM audience uploadsHostedSolo $20/mo; Scale $500/mo
Improvado Multi-sourceNot statedNot statedClaimed, not documentedAlertsPipeline-influence metricHostedDemo only
Supermetrics Multi-platformNot statedNot statedLinkedIn not confirmedNot statedNot statedHosted; Data API planNot on MCP page; 14-day trial
Two Minute Reports ReportingNot statedNot statedReads onlyNot statedNot statedOAuthFrom $97/mo ($81 annual)
Zapier MCP AutomationNot statedNot statedAudiences, reportsNoNot statedPersonal server URLFree 100 tasks; Pro from $29.99/mo
Adzviser Multi-source dataNot statedNot statedNone describedNot statedNot statedHostedFrom $48.99/mo
Adside AI ad agencyNot statedNot statedNot statedNot statedNot statedAPI key$990/brand/mo
Funnel Data hubNot statedNot statedReads onlyNot statedNot statedOfficial Claude, ChatGPT connectorFunnel customers
danielpopamd/linkedin-ads-mcp Open sourceNoNoYesNoNoSelf-host + own LinkedIn appFree (MIT)

Kiin built Kiin Intelligence and lists it first. Every other row comes from that vendor's own site, docs or repository, checked 26 Oct 2026. “Not stated” means we could not find it on their site, not that the feature is missing. Prices are as published and may be billed annually.

Also in the field, mostly reporting connectors and automation platforms:

ToolWhat its own site saysPricing as stated
DataslayerHosted MCP over its data connectors. Its docs say access is read-only.From €29/mo (annual)
Porter MetricsHosted MCP; read-only, no write scopes requested.Free tier for 3 ad accounts
Coupler.ioHosted MCP for querying and analysis; no writes stated.MCP on higher paid plans
CatchrHosted MCP for reporting; writes not stated.From $24/mo
Data BlooRead-only MCP and ChatGPT plugin.Free to $299.99/mo
GrowthSpreeClaude Desktop extension; recommends, you apply changes in Campaign Manager.Free to start
ComposioLinkedIn Ads toolkit: analytics, lead forms; no campaign create, edit or pause; bring your own LinkedIn app.Platform pricing
PipedreamLinkedIn Ads report actions and conversion events.Not stated
MakeRuns Make scenarios you build, including LinkedIn Ads modules.Make plans
jeptoLinkedIn Ads as a reporting source; KPI and budget checks.From $10/mo
CDataOpen-source read-only server on a licensed JDBC driver; hosted Connect AI for more.Licence; no list price
adwarpOpen source (MIT); drafts LinkedIn campaigns and pauses ads after asking; your own developer app.Free
Nuraveda/linkedin-ads-mcpOpen source; reads, creates draft groups and campaigns, changes status; self-host with your own app.Free (MIT)
mharnett/mcp-linkedin-adsOpen source; read-only tools today.Free (MIT)
radiateb2b/mcp-linkedin-adsOpen source, last updated June 2025; its product page now redirects elsewhere.Free

Full entries, with what each is best for, are in the best LinkedIn Ads MCP servers and AI connectors.

Is an MCP server better than Campaign Manager, an export or the raw API?

For ad hoc questions, yes. Campaign Manager answers the questions it was built for; exports need spreadsheet work; the raw API needs engineering. An MCP server lets you ask a follow-up in seconds. The raw API is still the right tool if you have engineers and want your own pipeline.

Campaign ManagerCSV exportRaw APIKiin MCP
Ask a follow-upRe-navigateRe-exportWrite more codeJust ask
Join data across viewsManualSpreadsheet workIf you build itBuilt in
Landing page metrics as primaryNoNoIf you know toEnforced
Make changesYesNoIf approved for writeYes, preview-then-confirm
Setup effortNoneNoneDeveloper app and weeksMinutes
BenchmarksNoneNoneNone1,000+ accounts

If you go the raw API route, three quirks catch almost everyone: monetary values arrive as strings, timestamps are Unix epoch milliseconds, and the analytics endpoint returns only impressions and clicks unless you name every other field. More in LinkedIn Ads API vs MCP and getting LinkedIn Ads data into PostgreSQL.

Is it safe to connect an AI assistant to a LinkedIn ad account?

It is as safe as the permissions and the confirmation step. Kiin sees only the ad accounts you authorise. Nothing changes without a confirmed second call, there is no autonomous process spending money in the background, and every write is logged.

  • Scoped access. Only the LinkedIn ad accounts you connect, plus any CRM you choose to connect.
  • Preview-then-confirm. The first write call shows the before/after diff and changes nothing.
  • Drafts. New campaigns are created as drafts; you activate them in Campaign Manager.
  • Guardrails. Audience Network and Audience Expansion stay off; no bid raises on fatigued campaigns.
  • Token health. Ask “is my LinkedIn connection healthy?” before you trust a number that looks stale.

Treat AI output as a draft, not a decision

An assistant querying live data is far more reliable than one guessing, but check anything that would move real budget before you confirm it.

How much does a LinkedIn Ads MCP server cost?

Kiin is free to start on one ad account, with a 14-day Analyst trial and no card. Kiin Analyst is $65 per LinkedIn ad account per month.

Rival pricing, as each vendor publishes it, is in the table above.

Glossary

MCP (Model Context Protocol): an open standard for connecting AI assistants to external systems.

Remote MCP server: a hosted server you add by URL. Kiin's is https://mcp.kiin.co, with /mcp for streamable HTTP and /sse for SSE-only clients.

OAuth 2.1 with dynamic client registration: how your assistant signs in to Kiin without a developer app or API key.

Landing page CTR: landing page clicks ÷ impressions. The traffic click-through rate for feed ads.

Cost per landing page click: spend ÷ landing page clicks.

Total CTR: all clicks ÷ impressions, including social engagement. Context only.

Influenced pipeline: deals at companies your LinkedIn ads reached or engaged before the deal.

Would rather talk it through? Book a call with the Kiin team →