Vector MCP: Overview

The Vector Model Context Protocol (MCP) server connects AI agents — Claude, ChatGPT, Cursor, and others — directly to your Vector account. Instead of pulling reports, exporting CSVs, or jumping between Vector and LinkedIn Campaign Manager, you ask questions in plain English and your agent pulls the answer.

Quick start: Setup guide for your AI client of choice. Setup takes about 2 minutes once you have your credentials.

What you can do

The Vector MCP currently gives your agent secure, read-only access to two data sources:

• Vector visitor data — first-party site intelligence: who visited your site, from where, what they read, ICP match, enriched person and company attributes.

• LinkedIn Ads data — ad platform metrics: spend, impressions, clicks, leads, pacing, creative performance, and company-level engagement.

• Google Ads data — coming soon

• Meta Ads data — coming soon

Once connected, you can ask things like:

• "Who from our top 50 target accounts visited the pricing page this week?"

• "How is our LinkedIn ad budget pacing this month? Flag anything underpacing."

• "Which creative is driving the most ICP traffic in the Q4 ABM campaign?"

• "Show me companies that engaged with our retargeting ads but never visited the site."

Your agent picks the right tools, runs the queries, and returns a structured answer. No SQL, no dashboards, no waiting on the data team.

Who it's for

How it works

┌─────────────────┐       ┌────────────────┐       ┌────────────────────┐
│  Your AI agent  │ ────▶ │  Vector MCP    │ ────▶ │  Vector + LinkedIn │
│ (Claude, etc.)  │       │  server        │       │  data layer        │
└─────────────────┘ ◀──── └────────────────┘ ◀──── └────────────────────┘
      natural               authenticated             your account data
      language               tool calls                 (read-only)

1. You ask a question in your AI client.

2. The agent picks the right MCP tool(s) for the job.

3. The MCP server authenticates as you and runs the query against your Vector tenant.

4. Results come back as structured data; the agent translates into a natural-language answer.

What's available today

The MCP exposes 12 tools across two surfaces:

Vector visitor tools (4)

• vector_visitors_search — find specific visitors or companies; filter by UTM, page, source, or campaign

• vector_visitor_traffic_breakdown — aggregate site traffic split by source, UTM, or company

• vector_top_pages — most-visited pages (overall or ICP-only)

• vector_campaign_traffic — ad-driven site traffic pivoted by campaign, source, or LinkedIn attribute

LinkedIn Ads tools (8)

• linkedin_performance_summary — compare performance across campaigns, groups, or accounts

• linkedin_campaign_details — deep dive on a single campaign

• linkedin_campaign_find — resolve a campaign name to an ID

• linkedin_performance_trend — time-series metrics at account, group, or campaign level

• linkedin_campaign_pacing — budget pacing, projected spend, pacing buckets

• linkedin_creative_performance — ranked creative performance within a campaign

• linkedin_engaging_companies — companies engaging with your ads (ABM signal)

• linkedin_company_campaigns — inverse drill-down: which campaigns a single company engaged with

Authentication

The Vector MCP uses OAuth 2.0. Your AI client opens a browser, you log in to Vector once, and a scoped access token is stored in your client. The token is tied to your Vector account — the MCP can only access data you already have permission to see in the Vector UI.

Tokens refresh automatically. If a token expires or is revoked, your client will prompt you to re-authenticate - this will happen every 14 days.

The MCP is read-only. It cannot modify campaigns, change settings, or write data back into Vector.

Pricing and access

The Vector MCP is included with all Vector plans at no additional cost. Every Vector user can authenticate and use it.

Vector MCP is currently in alpha - if you'd like to join our current customer cohort contact your CSM or support@vector.co.