“Which of my charge points faulted overnight, and why?”
An operations manager can ask their AI assistant a similar question on their phone by voice, while driving to the office. The assistant pulls the answer from live data: the charge points that faulted overnight, and what their logs show. Until now, that question waited until someone was back at a desk with the operator dashboard open.
The AMPECO MCP server connects Claude, ChatGPT, or another supported AI assistant to your AMPECO EV charging platform, so your team can ask about the network and get answers from live data.
Your AI assistant, connected to your charging network
If your team works in ChatGPT or Claude throughout the day, those assistants know nothing about your charge points, sessions, tariffs, or roaming partners on their own.
Model Context Protocol (MCP) is the open standard AI assistants use to connect to other software. The AMPECO MCP server is that connection for your AMPECO platform. Add it to Claude (on the web, on desktop, or in Claude Code), to ChatGPT, or to another assistant that supports the sign-in standard they use (CIMD). The assistant can then work with your live network data.
It uses the same login as you do for your AMPECO platform account – through your usual sign-in, the same one you use for the operator dashboard, and with the same rules, including single sign-on and two-factor authentication where your tenant uses them. Nobody creates an API key, a token, or a client secret, and no developer has to set anything up.
The MCP server provides the assistant with the platform’s business rules for each area (how tariffs apply, what a session status means, which settings depend on one another) and asks it to read them before it makes a change.
AMPECO’s APIs have long let operators connect the payment, software, and business tools they choose. The MCP server adds the AI assistant they choose to that list. Because it is built on the Public API, the MCP server grows with it, with no update on your side.
What you can now do
With the MCP server connected, you can ask your assistant to find any record you can see, answer reporting questions from data you have access to, explain failures from charge point and roaming logs, and make changes you confirm.
Look up any record you can see, with a link to it
“Which charge points at the Central Station location are offline right now? Give me links.”
The MCP server returns the matching charge points, each with a link that opens its record in the operator dashboard. It is built to strip passwords, tokens, and other credentials from what it returns, even when they are part of a record.
Answer reporting questions without an export
“How many sessions did each location run last month, and which hour of the day is busiest?”
The MCP server gives the assistant counts, totals, averages, and grouped figures. They are calculated on the platform, so they cover every record you can see, not one page of results.
Find out why a charge point faulted or a session stopped
“Why did the session on charge point 3137, connector 2, stop at 14:32 yesterday? Read the OCPP log.”
The MCP server gives the assistant the charge point’s OCPP (Open Charge Point Protocol) log, the messages exchanged between the charge point and the platform, so it can explain in plain words what happened. Ask “Which of my charge points had the most failures this week?” and it can compare failures across your charge points. Roaming problems work the same way, through the OCPI (Open Charge Point Interface) log with your roaming partners.
Your support team gets a first diagnosis without reading raw protocol messages, and a field technician knows the likely cause before driving out. For drivers, that means charge points back in service sooner.
Make a change, after you confirm it
“Move these three charge points to the Weekend tariff.”
The MCP server labels every operation as a read or a change, so assistants like Claude and ChatGPT can ask you to confirm before a change runs. How and when they ask is set in your assistant.
Changes run under your role in the platform, so your role’s restrictions apply to every change you confirm in your assistant. Changes also need a connection set to Read and change, which you choose when you sign in (step 3 below). If you want no changes at all, keep the connection on Read only, where the platform itself refuses them.
Connect it in three steps
- Copy your MCP server URL. In the operator dashboard, open My account, go to MCP and copy the URL shown there.
- Add it to your assistant. In Claude, ChatGPT, or another supported assistant, add the URL as a custom connector.
- Sign in and choose access. Log in to your AMPECO platform account and choose Read only or Read and change. The sign-in screen displays your brand and the name of the assistant requesting access. Which one you pick decides what the assistant can do through the MCP server.
In company workspaces, your assistant’s own admin settings may decide who can add custom connectors. Each person still signs in with their own AMPECO platform account.
You decide what it can see and do
The MCP server acts as you: every request runs under your account, with your role and operator scope. If your account is restricted to an IP allowlist, the assistant’s servers must be added to it.
Read only is the default. With Read only, the assistant can look things up, answer reporting questions, and explain failures, and the platform itself refuses any change. Changes are possible only with Read and change. Access never widens on its own: to change your choice, you should disconnect and reconnect.
You can disconnect an assistant from your account at any time, and operator admins can revoke any connection within their organization. Deactivating an admin or removing their MCP permission ends access on the next request.
The MCP server and your AI assistant are separate systems. AMPECO operates the MCP server: it controls what the server returns, under whose permissions, and which changes it accepts. The MCP server uses no AI model, and AMPECO sends no data to any AI provider through it. Your organization chooses and contracts your AI assistant. Data returned to it is processed by its provider under your agreement with that provider, so use an account and plan your organization has approved. CoOperator, the AI agent built into the platform, is a separate service under the CoOperator Terms of Service.
Answers and suggested changes depend on the assistant and model you use, so review any changes before you confirm them.
CoOperator and your own assistant: two ways in
CoOperator is the AI agent built into the AMPECO platform. It works in the sidebar of the operator dashboard, takes into account the page you are on, and carries context from one conversation to the next (what’s new in CoOperator).
The MCP server is for Claude, ChatGPT, or another assistant your team uses: at a desk, on a phone, by voice. The two work independently, and neither replaces the other. A support agent investigating a session in the operator dashboard has CoOperator beside them. An operations manager preparing a board update in Claude, or asking about the network on the way to a site, has the MCP server.
Ask your first question this week
If you already run on AMPECO, your account delivery manager can walk you through the setup. If you are evaluating AMPECO, book a demo.