The share of companies employing GTM engineers grew from 0.1% in January 2023 to 2.3% by June 2026, according to data tracked across roughly 9,600 companies by GTM Engineer School. That's a 23x increase in three and a half years. The reason is straightforward: GTM stacks became too fragmented for manual operation, and the glue code holding them together kept breaking.
Emily Kramer's MKT1 MCP Showcase, held on July 28, 2026, highlighted this trend. Twenty-one workflows across seven GTM tools ran through Claude via MCP (Model Context Protocol). The thesis: connect enough of your stack to Claude, and the AI becomes an operational layer rather than just a chatbot.
The Workflows That Matter
Kramer showcased seven standout workflows for operators focused on pipeline impact.
Attio MCP for CRM-driven event targeting. Kramer's team uses this to pull guest lists based on attributes of previous dinner attendees. This signal-based audience building occurs inside Claude, utilizing live CRM data. For demand gen teams running ABM dinners or field events, this replaces the manual export-filter-import cycle that consumes two to three hours per event.
Mutiny MCP for ABM landing pages. This workflow creates on-brand, engagement-tracked landing pages for target accounts without relying on PDFs. Jaleh Rezaei, Mutiny's CEO, remarked, "No more PDFs please." The measurement aspect is crucial; these pages automatically track engagement, providing account-level signals back into your pipeline data instead of a black hole.
Airtable MCP for cross-team request management. Marketing teams often manage requests via Slack, email, and meetings. The Airtable MCP scrapes requests from meeting notes or allows submissions through Claude directly. This workflow reduces friction between marketing and sales or product teams without requiring anyone to learn a new tool.
Zapier's MCP connector is noteworthy for its breadth, bridging Claude to over 9,000 apps, allowing any tool without its own MCP server to participate. This serves as a fallback, not a strategy.
Where Operators Should Start
Implementation guides and expert recommendations consistently advise against connecting everything. Start with two or three MCP servers that eliminate the most copy/paste work, beginning with read-only integrations.
The logic is straightforward. Read-only workflows (pulling CRM data, surfacing analytics, retrieving account intelligence) allow you to prove value and test permissions without risking data corruption. Daniel Kravtsov commented on Kramer's post, highlighting the permissions issue: "We ended up with a default deny gateway because individual tool permission models simply never agree."
This is the operational risk. Each tool in your GTM stack has its own permission model. When integrated into Claude via MCP, those permissions must carry through consistently. Softr's MCP demo showed this working correctly, but across a full stack, a governance layer is essential.
The phased approach looks like this:
- Phase 1 (read-only): CRM queries, analytics pulls, account research packs. Validate with real queries before expanding access beyond one or two power users.
- Phase 2 (limited write): Landing page creation, request intake, content updates. Scope tightly. Sandbox test. Set explicit OAuth and permission controls.
- Phase 3 (action-taking): Signal-based outbound triggers, automated routing, campaign creation. Only after Phases 1 and 2 have run without incident.
The Partial-Automation Gap
The relevance of MCP extends beyond demos. According to 2023 B2B automation data, 76% of businesses use marketing automation, yet 59% still operate with only partial automation. This gap is where manual glue work exists: the copy-paste between CRM and spreadsheets, the Slack threads asking sales for context, and meetings where data is read off dashboards.
MCP doesn't replace your tools; it reduces manual routing and research work between them. The value proposition isn't "AI does your job"; it's that Claude can query your CRM, pull analytics, and surface account context in one interface, eliminating the need for manual data handling across multiple tabs.
The market is evolving rapidly. In July 2026 alone, Meta expanded access to its ads MCP server, Snap completed its ads MCP rollout, Webflow launched MCP 2.0 with governance features, and Guideline introduced an ad intelligence MCP server. By August, Dash Social, Getty Images, and Decile announced MCP capabilities. The infrastructure layer is filling in quickly.
However, vendor launches aren't the main focus. Frank Howard's comment on Kramer's showcase post captures the operator's reality: "Demos are great, but actual user feedback—like how teams adapt and tweak these workflows—can reveal hidden value or pitfalls. The real power lies in iteration, not just setup."
He's right. The teams that derive value from MCP won't be those that connect the most tools but those that select two workflows, run them for a month, measure changes in handoff speed and data accuracy, and then decide what to wire up next. The showcase featured 21 workflows; your rollout should start with two.
