your AI configures heysteff for you.

MCP (Model Context Protocol) is an open standard created by Anthropic so an AI assistant can connect to external tools and data in a consistent way, instead of every piece of software inventing its own way to "talk" to an AI. In practice, it's a shared language: if an app exposes an MCP server, any AI client that understands the standard can read its data and take action there, with no custom integration required.

heysteff exposes exactly that: a remote MCP server that represents everything you'd normally do from the dashboard, available for your own AI to operate.

With heysteff's MCP server connected, your AI can read and configure nearly everything a user would do from the web app: the knowledge base that trains the agent, Steff's behavior and personality, automation flows, the custom tools that connect to your API or webhooks, and your connected channels. It's the difference between asking someone on your team to open the dashboard and make a change, and just asking your AI in a conversation.

This is especially useful if you already use an AI to manage your day-to-day operation: instead of that AI just "knowing about" heysteff, it can actually operate it — update a policy in the knowledge base, adjust a flow, add a new tool — all inside the same workflow you already live in.

heysteff's MCP server is remote (nothing to install locally) and authorizes through OAuth: you approve access from your heysteff account, the same way you'd connect any other app to your Google or Shopify account. It works with Claude and, more broadly, with any client that implements the MCP standard — it isn't tied to one specific tool.

Giving an AI the ability to configure your support software sounds like a risk without visibility. That's why every write action your AI takes through the MCP server — changing a rule, creating a flow, touching a tool — is logged in the same audit log the workspace owner already uses to see what changed, when, and through which channel. Nothing happens silently.

It's not a feature every tool in the category offers today, and we say that without overselling it — it's a narrower, more specific capability than the core product, but a real one. If the way you already work runs through an AI assistant connected to your tools, being able to say "update the refund policy in heysteff" or "create a new flow for this case" and have it actually happen — with the same audit trail as any other change — is a genuinely different way to run your support, closer to how you already work with AI the rest of your day.

Also relevant: copilot mode, appointment scheduling, AI lead qualification, AI agent for ecommerce and servidor MCP (Spanish).