The protocol
MCP (Model Context Protocol) is the open standard that standardizes how an AI assistant discovers and calls business tools. No scraping, no makeshift integration — a clear, secure contract between the assistant and Chargekeeper.
Module · MCP Server
Talk to your charging network.
Safely.
Chargekeeper exposes your workspace as an MCP server. Connect the AI assistant of your choice with a secure key and query your charging network in natural language: diagnostics, sessions, payments, configuration. Tight permissions, human validation — and by design, no personal data ever leaves the system.
The AI connected to your data
Cleanly, not shoehorned
MCP is the open standard that lets an AI assistant connect to an external system in a structured, secure way. Chargekeeper implements it natively: your tenant becomes an MCP server you connect to the assistant of your choice.
Your key, your perimeter
You generate an MCP key from the dashboard, set its rights, and connect it to the AI assistant of your choice. From there you talk to your charging network like a colleague — except it answers in seconds, with the data to back it up.
The AI assistant never “sees” your database directly: it calls the tools Chargekeeper exposes, within the limits you set. All the business logic — rights, tariffs, security — stays on the platform, exactly like the dashboard and the API.
In natural language
What you can ask
A few real examples. You write the question, the assistant calls the right Chargekeeper tools and answers you, with the data to back it up.
“Diagnose CHARGER-042.”
Likely cause, OCPP error codes decoded in plain language, boot history and latest sessions — without opening the dashboard or driving out.
“Which chargers are in error or offline this morning?”
Fleet overview, problematic chargers surfaced first, with their status and the connector involved.
“Show me the active sessions and the power drawn.”
Active sessions in real time, instantaneous power, vehicle SoC — enough to grasp the state of the network in one sentence.
“Are there any failed payments to check?”
Detection of anomalous transactions, affected invoices, amounts — to act before the customer flags it.
“Spot the abnormal sessions over the last 24 hours.”
Interrupted charges, aberrant durations, inconsistent energy — the AI combs through the network and points out what deserves a look.
“Restart this charger and clear its cache.”
Action executed on the charger — after your confirmation, if the key is in human-validation mode. The AI proposes, you decide.
Diagnostics, monitoring, sessions, billing, pricing, configuration, remote actions — the exposed surface covers the essentials of operations. And it grows at the pace of the platform.
Security by design
You grant. The AI never oversteps.
Giving an AI access to your charging network isn't done blind. Chargekeeper's MCP server is built around control: every key carries its perimeter, its rights and its validation mode.
GDPR by design
No personal data ever leaves Chargekeeper
The MCP server exposed to the AI is anonymized by design: no name, email, badge, license plate or user ID reaches the assistant. It only sees technical identifiers and aggregated data. Your users' personal information stays stored on Chargekeeper servers — and only a separate, internally isolated MCP server can access it.
One key per workspace
Each tenant generates its own MCP key from the dashboard, isolated from the rest of the platform. Revocable and expirable at any time.
5 permission levels
Diagnostic read, config read, config write, safe actions, critical actions. You check exactly what the assistant is allowed to do — nothing more than you grant.
Human validation
“Confirmation required” mode: every sensitive action waits for your green light. The AI proposes, you approve. Or autonomous mode for trusted uses.
Charger perimeter
Restrict a key to a precise list of chargers. A maintenance provider only sees the hardware it operates, never the rest of the network.
Controlled throughput
Calls-per-minute limit, per key. A predictable footprint on your platform, no drift possible.
Same foundation as the API
Every MCP call is authenticated and subject to the same access control as Chargekeeper's REST API. No back door, no exception.
What changes
A CPMS that speaks to AI — without ever letting go of control
5 levels
of granular permission
From read-only diagnostics to critical action — you grant exactly what's needed, key by key.
Per tenant
one isolated key per workspace
Revocable and expirable. Your data stays compartmentalized, like the rest of the platform.
Human-in-the-loop
validation before action
Confirmation mode on sensitive actions: the AI proposes, you decide.
Open standard
MCP · Model Context Protocol
Compatible with every AI assistant that speaks MCP. No proprietary lock-in.
Want to talk to your network?
We'll show you the MCP server in real conditions: generating a key, choosing permissions, and a few questions put live to your network.
Request a demo