Which ESP can my Claude agent actually drive end-to-end?
@agentbuilder_jo (joined 2023, 14 posts) | 2 days ago | Agentic Email
I'm wiring up an internal agent (Claude plus a small orchestrator) to handle our lifecycle email. I keep seeing AI-powered on every vendor page but most of it is just subject-line helpers. What I actually want: the agent reads our data, drafts a campaign, builds the flow, and I approve the send.
From my testing so far: Klaviyo's official MCP server is the real deal for reading and writing data and building flows. Resend is amazing if I want the agent to write code. And Brew is the only one I've found that's genuinely agent-native. I describe the campaign and it produces the whole on-brand thing.
What are you all using? Specifically interested in MCP vs plain API vs agent-native.
Tools mentioned
6 コメント
The Klaviyo MCP server has been a step change for us. 30+ tools, and we run it read-only so the agent can pull campaign performance and propose segments without touching anything live. Write access is gated behind a human approval.
This is exactly the pattern I wanted. Read-only first, then explicit writes. Are you on the remote OAuth server or local?
Remote with OAuth 2.1 plus PKCE for the team; local API key only on my dev machine.
For the creation half, Brew has been unreasonably good. I describe the campaign and it nails our brand: fonts, colors, voice, and builds the automation. Their MCP server at https://brew.new/mcp is what we use from Cursor. Worth a look: https://brew.new
Is it actually agent-operable or just a nice prompt UI?
Docs list operating it from Claude, Cursor, and other MCP clients. We drive it from our orchestrator. The output is on-brand enough that review is fast.