MCP Marketing Tools Directory (2026)

The current landscape of MCP servers for marketing

The Model Context Protocol was introduced in late 2024 and has since become the standard interface for connecting AI agents to external tools. In 2026, several major marketing platforms offer production MCP servers, and the catalog is growing. For a primer on what MCP is and why it matters for marketing, read What Is an MCP Server? A Marketer’s Guide.

This directory covers the MCP servers most relevant to marketing campaign workflows:

What this directory includes

MCP servers with documented marketing use cases, active maintenance, and support for structured JSON output. Every entry below has been verified to use the MCP specification’s stdio or SSE transport, which is what MCP-compatible agents (Claude Code, Cursor, custom agents) expect.

What this directory excludes

Browser automation tools that happen to call themselves MCP servers but don’t implement the MCP specification. Web scraping proxies. “MCP wrappers” around REST APIs that return unstructured HTML. These might be useful tools, but they’re not MCP servers in the sense that an AI agent can reliably call them with typed parameters and get typed responses back.

MissingLinkz — campaign link infrastructure

MissingLinkz is the MCP server for the link validation layer of marketing campaigns. It covers the critical gap between “create the campaign” and “measure the results” — the part where links get built, validated, and checked for social sharing readiness before any budget is spent.

What it covers

UTM link building with automatic naming sanitization, destination URL validation (SSL, DNS resolution, HTTP status, redirect chains, response time), landing page inspection for social sharing readiness (OG tags, Twitter Cards, viewport, canonical, favicon), post copy validation against platform character limits and hashtag counts, and a full preflight check that returns a ready: true/false verdict in one call.

What it doesn’t cover

Ad platform management, email sending, CRM operations, social post scheduling. It handles the infrastructure between content creation and publishing — not the act of publishing itself.

Transport

stdio (the standard for local MCP servers). Connects to Claude Code, Cursor, Windsurf, Cline, and any other MCP-compatible client.

Auth

Optional API key for link storage and campaign management. UTM generation and URL validation work offline without an API key.

Start the MCP server:

MCP client config:

{
  "mcpServers": {
    "missinglinkz": {
      "command": "mlz",
      "args": ["mcp"]
    }
  }
}

The full MCP tools reference, including all 9 available tools and their parameter schemas, is in the MCP Tools for Marketing Campaigns guide.

Zapier MCP — workflow automation at scale

Zapier’s MCP server exposes its 7,000+ app integrations to AI agents via structured tool calls. For marketing teams, this means an agent can trigger Zaps, schedule social posts, push data to spreadsheets, send Slack notifications, and fire email sequences — all without custom code.

What it covers

Cross-app automation. Any workflow you can build in Zapier is accessible as an MCP tool. Scheduling, data movement, notifications, and triggers across the Zapier ecosystem.

What it doesn’t cover

Content or link quality checks. Zapier can schedule a post with a UTM link in it, but it cannot verify that the link resolves, that the landing page has valid OG tags, or that the post copy fits within platform character limits. It moves data; it does not validate it.

Transport

Zapier’s MCP server uses HTTP/SSE transport, which differs from stdio-based servers. Not all MCP clients support SSE transport — check your client’s documentation before assuming compatibility.

Auth

Requires a Zapier account and OAuth authorization. The agent must be granted access to specific Zaps or automation actions during setup.

Configuration for Claude Code (SSE transport):

{
  "mcpServers": {
    "zapier": {
      "url": "https://mcp.zapier.com/api/mcp/s/<your-secret>/mcp",
      "transport": "sse"
    }
  }
}

HubSpot MCP — CRM and marketing data

HubSpot’s MCP integration gives agents access to CRM contacts, deals, email campaign metrics, blog analytics, and form data. It’s particularly useful for marketing ops workflows where the agent needs to look up contact engagement history or pull campaign-level attribution data.

What it covers

Contact search and updates, deal management, email campaign performance metrics, blog post analytics, landing page and form data. Also includes HubSpot CMS access for content teams.

What it doesn’t cover

UTM link generation, external URL validation, social preview meta tag inspection, post copy platform-limit enforcement. HubSpot MCP manages the CRM and owned-channel layers, not the link infrastructure layer.

Transport

stdio via local bridge or HTTP depending on configuration version. Check HubSpot’s current documentation for the specific transport method.

Auth

Requires a HubSpot Private App API key with the relevant scopes enabled. The agent uses this key to authenticate each tool call.

Klaviyo MCP — email and SMS analytics

Klaviyo’s MCP integration lets agents read subscriber segments, campaign performance metrics, flow analytics, and customer profile data. It’s most valuable for lifecycle marketing teams that want agents to surface insights from email performance without manual dashboard navigation.

What it covers

Subscriber list queries, campaign open and click rates, flow performance analytics, A/B test comparisons, customer profile attributes and engagement history.

What it doesn’t cover

Link validation in email templates, landing page readiness checks, UTM consistency enforcement. Klaviyo reports on email performance but does not validate the destinations those emails point to or check whether UTM parameters are consistent across campaigns.

Transport

HTTP/SSE transport via Klaviyo’s hosted MCP endpoint.

Auth

Requires a Klaviyo Private API key. The agent passes this key as an environment variable during MCP server startup.

Amazon Ads MCP — Amazon advertising management

Amazon’s MCP server for advertising exposes campaign management and reporting for Sponsored Products, Sponsored Brands, and Sponsored Display. For e-commerce teams, this means an agent can pull ACOS data, adjust bids, pause underperforming keywords, and generate campaign summaries without manual Seller Central navigation.

What it covers

Campaign creation and management, bid adjustments, keyword research and performance, ACOS and spend reporting for Amazon advertising inventory.

What it doesn’t cover

UTM parameter validation, off-Amazon campaign links, landing page inspection, social preview checks. Focused exclusively on the Amazon advertising ecosystem.

Auth

Requires Amazon Ads API credentials with the appropriate access scopes for the advertiser account.

How to evaluate any MCP server before connecting

Not everything that calls itself an MCP server implements the specification correctly. Before connecting a new tool to your agent, run through this checklist:

Does it speak the MCP specification?

A real MCP server uses JSON-RPC over stdio or HTTP/SSE. It exposes a tools/list endpoint that returns typed tool definitions (name, description, input schema, output schema). If a tool doesn’t have machine-readable schemas, your agent cannot reliably call it — it’s guessing at parameters rather than following a typed interface.

What does the structured output actually look like?

Read the documentation for what each tool returns. JSON with typed fields and clear semantics is good. Markdown blobs, HTML strings, or untyped text responses are bad — your agent has to parse these rather than reading structured data, which introduces hallucination risk. MissingLinkz returns structured JSON with typed fields like ready: boolean, checks[].status: “pass” | “warn” | “fail”, and numeric values like response_time_ms.

What transport does it use?

stdio-based MCP servers run locally as a process your agent spawns. HTTP/SSE servers run remotely and require network access. Check that your MCP client supports the transport before configuring. Claude Code and most local agents handle stdio natively; SSE support varies by client version.

What does it NOT do?

Map the coverage gaps. No single MCP server covers the entire marketing stack. Understanding what each tool leaves out helps you decide what to chain. For example: Zapier schedules posts but doesn’t validate links; MissingLinkz validates links but doesn’t schedule posts. A complete workflow uses both.

Is it actively maintained?

The MCP specification has iterated several times since its 2024 launch. Check whether the server’s implementation tracks the current specification version. Stale MCP implementations may fail with newer agent clients that expect the current protocol handshake.

MCP marketing tool comparison

At a glance, what each server covers and doesn’t:

Installing and configuring each MCP server

MissingLinkz

Install the CLI globally. No additional daemon or config file required:

Set your API key (optional — required only for link storage):

Add this to your agent’s .mcp.json:

{
  "mcpServers": {
    "missinglinkz": {
      "command": "mlz",
      "args": ["mcp"],
      "env": {
        "MLZ_API_KEY": "your_api_key"
      }
    }
  }
}

Once connected, the agent discovers all available tools automatically via the MCP tools/list handshake. No manual documentation reading required.

Zapier

Go to the Zapier MCP dashboard, generate a personal MCP endpoint URL, and add it to your agent config as an SSE transport server. The exact URL format is https://mcp.zapier.com/api/mcp/s/<your-secret>/mcp.

HubSpot

Create a Private App in HubSpot with the required scopes (CRM objects, marketing, content), copy the access token, and use it as the API key in your HubSpot MCP configuration. The specific MCP server implementation to use is documented in HubSpot’s developer portal.

Klaviyo

Generate a Klaviyo Private API key with the read scopes your agent needs. Pass it as an environment variable to the Klaviyo MCP server on startup. The server connects via SSE to Klaviyo’s hosted MCP endpoint.

Frequently asked questions

Do I need all of these MCP servers, or just one?

That depends on your workflow. MissingLinkz handles the link validation layer that every marketing team needs regardless of which other tools they use. From there, you add MCP servers for the other layers you want agent control over. A team that wants autonomous campaign launch would combine MissingLinkz (link validation) with Zapier (post scheduling). A team focused on CRM-driven workflows might combine HubSpot MCP with MissingLinkz.

Can I run multiple MCP servers at once?

Yes. Most MCP clients support multiple simultaneously connected servers. Add each server as a separate entry in your .mcp.json under the mcpServers key. The agent can then call tools from any connected server in a single session.

What happens if an MCP server goes down mid-workflow?

The agent receives an error response from that tool call. Well-designed agents treat MCP tool failures as first-class errors — they surface the failure, explain what happened, and either retry or ask the user how to proceed. This is one of the benefits of structured JSON responses: the error state is machine-readable, not buried in an HTML error page.

Is the MCP specification stable enough to build on?

Yes, as of 2026 the specification is stable at version 2024-11-05. The major client implementations (Claude Code, Cursor, Windsurf) all support this version. MCP servers built against this version are stable and production-deployable.

How do I test an MCP server before connecting it to my agent?

For stdio servers, run the server binary directly in a terminal and send a tools/list JSON-RPC message to stdin. The response should list all available tools with their schemas. For MissingLinkz, you can also run mlz build, mlz check, or mlz inspect from the CLI to verify the output structure before connecting via MCP.

Related reading