Enterprise marketing teams use Knak to build emails with brand governance, approval workflows, and rendering that works across every email client. That production infrastructure has always been accessible through the visual builder and the Knak API. With the Knak MCP Server, it's now accessible to your AI assistant too.
Model Context Protocol is the open standard that's becoming the default way AI tools interact with external systems, and the Knak MCP Server uses it to connect tools like Claude and ChatGPT directly to your Knak instance. From your AI workflow, you can browse your brands, campaigns, and themes, and generate email assets inside Knak without switching to the platform.
This is an alpha release, so it's worth being clear about scope. The MCP Server today covers workspace discovery and email generation, and we're building on that foundation. What matters more than the current feature set is what this represents: Knak's production infrastructure becoming accessible to a new type of operator. The platform runs the same whether a human is building in the visual editor or an AI agent is working through MCP. Same rendering pipeline, same brand controls, same production quality. The interface changes, but the infrastructure doesn't.
What the Knak MCP Server does today
The MCP Server gives your AI assistant access to tools that cover workspace discovery and email generation.
Watch the Knak MCP in action!
Explore your workspace
Your AI assistant can browse the brands, campaigns, and themes in your Knak instance. This means you can ask questions like "what brands do I have access to?" or "show me the campaigns in this brand" and get answers directly from your Knak data, without opening the platform.
- List brands accessible to your account
- List campaigns within a brand
- List themes available for generation, filterable by name or published status
- Search existing assets by brand, campaign, name, or type
- Get asset details for any asset by ID
Generate email assets
Describe the email you need in a natural language prompt, specify the brand and campaign, and Knak generates the asset inside your instance. You get a link to view it in Knak Studio immediately, where you review and refine using the same tools and workflow you already know. You can optionally specify a theme, subject line, sender details, and tags.
The generated asset runs through the same rendering pipeline as any Knak email. The hard problems of email rendering, the Outlook compatibility, dark mode handling, responsive behavior, and CSS management that trip up LLMs working with raw HTML, are handled by the infrastructure underneath.
What customers are already building with Knak and AI
The MCP Server is one part of a broader shift happening across our customer base. Enterprise teams are building AI-driven workflows that use Knak as the production layer in the middle:
- OpenAI intakes campaign requests from marketers through Slack, structures them with a Codex Agent into Linear tickets, generates emails through Knak, then the marketer refines in the editor before deployment. Slack message in, brand-compliant email out.
- Meta connects an internal campaign brief tool (Launchpad) to Knak through API integration. AI generates the content, Knak assembles the asset with branded content, brand controls and approval routing, humans review and ship.
- Stripe built a custom "validation agent" that connects to Knak at the QA step, automatically validates whether final Knak email matches the original marketing brief requirements before they moving to deployment.
- Google uses Knak as the creation node at the center of every campaign, connected to their internally built AI orchestrator and tools both upstream and downstream.
- Nvidia runs a fully human-driven workflow from Figma design through Knak production to their custom Viva Translate AI tool for localization, then deploys to Marketo and SFMC.
These teams aren't waiting for one tool to do everything. They're building workflows where each system handles what it does best, and the connections between them make the whole thing work. The Knak MCP Server makes those connections simpler and more standardized.
How MCP fits alongside other tools in your stack
MCP servers are designed to work together. The Knak MCP Server handles email production. Other servers handle the rest of the workflow:
- Your CRM MCP gives your AI access to contact data and deal context
- Your analytics MCP gives your AI access to performance data
- Your project management MCP gives your AI access to campaign briefs and task tracking
- Your design MCP gives your AI access to brand assets and design context
With multiple MCP connections, your AI assistant can pull a brief from your project management tool, reference performance data from your analytics platform, and generate an email in Knak, all from one conversation. Each server provides a piece of the context. Together, they give your AI operator the full picture.
This is the direction enterprise marketing workflows are heading. One standard for connectivity means you're not locked into a single AI model or a single vendor's ecosystem. As long as the tool supports MCP, it works with Knak.
How to connect in ChatGPT
The connection takes about two minutes. In ChatGPT, open the Apps menu in the left sidebar and look for the Knak app. Click on it, then click Connect, and follow the prompt to continue to Knak for authentication. Once the connection is established, Knak's MCP tools are available in your chat.
You can start a new conversation directly from the Knak app, which pre-loads the MCP connection automatically. If you're already in a conversation and want to bring Knak in, type @ in the prompt field and select Knak from the list. From there, you can prompt normally. A prompt like "create an email promoting our upcoming webinar in the [brand name] brand under the [campaign name] campaign" is enough to get started, and the MCP server handles the rest.
You do need to tell the server where to create the asset by specifying a brand and campaign, since the MCP server needs a location in your Knak instance to place the generated email.
How to connect in Claude
In Claude, click the briefcase icon to open the customization panel, then click on Connectors. Search for "Knak" and you'll see Knak Enterprise Custom appear in the results. Click Connect, then Allow to authorize the connection. Claude will show you the full list of tools available through the MCP server once connected: generate asset, create preview, list assets, list campaigns, and more.
Once connected, you can start prompting right away in any conversation. The same prompt structure applies: describe the email you want, specify the brand and campaign, and the MCP server generates the asset inside your Knak instance while you continue working.
Access and prerequisites
The Knak MCP Server is available to all Knak Enterprise customers as part of the platform. You'll need an active Knak account with access to at least one brand, MCP enabled for your organization (your Knak admin handles this in company preferences), and access to either Claude or ChatGPT.
Authentication uses OAuth 2.1 with PKCE, but the AI clients handle the flow for you. You sign in with your Knak credentials when prompted, and the client stores the token for subsequent sessions. Your data access is scoped to your account permissions, so you'll only see the brands, campaigns, and assets you have access to in Knak.
What's coming next
This alpha release covers workspace discovery and email generation. It's the starting point, and we're building from here.
As we expand the MCP Server's capabilities, the goal is to progressively open up more of what Knak's production infrastructure already supports. The roadmap includes deeper integration with the platform's creation and governance features, building toward a future where AI operators have access to the same depth of functionality available through the visual builder.
We're also investing in Knak AI itself by carefully selecting the right model to ensure layouts follow the expected patterns for different types of emails with fewer obvious issues.
