Your AI agent can run Benson now
Benson's MCP server lets Claude, ChatGPT, Cursor and other agents set up a store, read its numbers and make changes you approve first.
Jump to a section
In short
Benson now has a remote MCP server. Connect Claude, ChatGPT, Cursor, VS Code, Claude Code, Codex, Gemini CLI or any other MCP client, and your agent can set up a store, answer questions about it and make the everyday changes. Every change shows you a preview first and waits for your yes.
Why build an MCP server?#
Because that's where a lot of store owners already work. They plan in Claude and ChatGPT. Agencies and developers spend their day in Cursor, VS Code and the terminal. Until now, getting a number out of Benson meant leaving that conversation, opening the dashboard and finding the right page.
MCP, the Model Context Protocol, is the open standard these tools use to connect to other software. With Benson connected, "How much did coupon extensions cost us last month?" gets an answer from your own data, in the conversation you're already having.
What can your agent do with it?#
Three kinds of work.
Set up. Give it your store's address and it can add the store to Benson, get the install snippet and the steps for your platform, check the storefront to confirm Benson is live, and apply the setup Benson recommends. It can also start a test session that shows Benson catching a simulated coupon pop-up on your own store.
Read. It can read most of what the dashboard shows: extension traffic and what it costs, the store audit, orders and the codes used on them, what Benson has protected, AI shopping agents on the store, affiliate overwrites, live activity, insights, your blocker and checkout settings, offers, experiments, your team, playbooks and our guides.
Change. It can switch the blocker between observe and block, start or stop a holdout, add and edit block rules, change your checkout code policy, draft and publish offers, act on insights and invite teammates.
There are also three ready-made prompts: set up a store, a weekly review, and stopping coupon pop-ups with a holdout to prove the change. The MCP server page lists every tool.
How do changes stay safe?#
An agent that can change your store needs firm limits. These are Benson's.
Every change asks first. When your agent asks Benson to change something, the first answer changes nothing. It's a preview: what would change, on which store, from what to what, and who it affects. Your agent shows you the preview and only goes ahead when you say yes. The yes travels as a single-use token that works for that exact change, once, within ten minutes. If anything about the request is different, it doesn't count.
Some things never go through an agent. Billing, deleting a store or its data, rotating keys, removing members, changing roles and exporting all your data stay in the dashboard, behind its own confirmations.
It can't do more than you can. Every call goes through the same checks as the dashboard: your role, your plan and your organization. Members can only read. If a change needs a plan you don't have, the preview says so before you're asked anything.
You can see it and stop it. Settings → AI agents lists every connection, what it can do, when it was last used and its recent tool calls. Disconnect stops it on its next call. Changes made through an agent show in your audit log as made via that client, so "via Claude" sits next to the name of the person who said yes.
Text from shoppers stays data. Some of what Benson reports was typed by someone else: UTM tags, campaign names, discount codes. Benson marks it as untrusted and tells your agent to treat it as data, not instructions. That lowers the risk of a page or a code trying to steer your agent. It can't remove it, which is one more reason changes wait for you.
How do you connect it?#
Add one URL to your agent and sign in to Benson. There are no keys to copy. You pick one organization for the connection and choose read-only, or read and changes. Agencies connect once for each client.
The MCP server page has one-click installs for Claude, Cursor and VS Code, and the steps for ChatGPT, Claude Code, Codex, Gemini CLI and any other client that speaks MCP. For agents that run without a person, like scheduled jobs and scripts, you can create an agent key instead. It's shown once, works in one organization and can be revoked any time.
Developers will find the details in the MCP server reference: discovery and scopes, every tool and the role it needs, the confirmation protocol, errors and limits.
It works on every plan, Free included. What your agent can do follows what your plan includes.
What's next?#
We're preparing listings in the Claude connector directory, ChatGPT's apps and the official MCP Registry, so you'll be able to find Benson from inside your agent. We'll add tools as the dashboard grows, and each one will follow the same rules: reads change nothing, and changes wait for your yes.
If there's something you'd like your agent to do in Benson that it can't do yet, tell us at [email protected].
