Use case

Add auth and email via Claude Code

AscendKit ships an MCP server covering auth, platform, content, surveys, journeys, email, and webhooks, plus AI-readable llms.txt docs. In Claude Code, Cursor, or any MCP client, you can configure your project conversationally instead of clicking through a dashboard, matching AscendKit's AI-conversational-first principle.

How it works

1. Connect the MCP server
Add the AscendKit MCP server to your AI coding client so it can read and configure your project.
2. Describe what you want
Ask the agent to set up auth, an email template, or a journey; it calls the matching MCP tools.
3. Review in the portal
Confirm the generated configuration in the dashboard, since sensitive credentials still use browser auth.

Why does an MCP server matter for setup?

Because configuration is the slow part, not the code. Adding auth means creating a project, choosing providers, verifying a sending domain, and authoring templates — most of it clicking through a dashboard while an agent waits.

AscendKit exposes those operations through an MCP server, so an agent can do them directly: create the project, enable providers, check domain verification status, create a template. The agent stays in one loop instead of stopping to tell you to go and click something.

The same operations are available through the CLI, which matters more than it sounds. An agent that can run a command can drive AscendKit even without MCP support, and you can reproduce anything the agent did by reading its commands.

Three ways to configure the same thing.
Surface	Best for	Reproducible
MCP server	Agents in Claude Code or Cursor	Yes
CLI	Scripts, CI, and agents without MCP	Yes
Dashboard	Exploring and one-off changes	No

What should you not delegate?

Anything involving a credential. AscendKit deliberately routes sensitive auth through the browser rather than accepting secrets as CLI arguments or chat input, because a secret pasted into a conversation is a secret in a transcript, a log, and possibly a model provider's storage.

Domain verification is the other one to keep human. It changes DNS records for a domain you own, and a mistake there affects your email deliverability well beyond AscendKit.

Everything else — project creation, provider configuration, template authoring, journey definitions — is safe to hand over, reversible, and visible in the dashboard afterwards.

FAQ

Which AI clients work with AscendKit's MCP server?

Any MCP-compatible client, including Claude Code and Cursor. The server exposes auth, content, surveys, journeys, email, and webhook tools.

Are credentials handled safely over MCP?

Yes. Sensitive credentials use browser auth rather than being passed as chat input, in line with AscendKit's security conventions.

Related guides

Add auth and email to Next.js — what the agent is setting up, written out by hand.
CLI reference — the same operations as commands, for agents without MCP.

Start with one API key

Auth, email, surveys, and journeys share one user record, so you ship this without stitching vendors together.

Start free