Which charge point management platforms have an MCP server?

As of 29 September 2026, all three compared here: askacharge.com (remote at https://askacharge.com/askacharge/api/mcp, 25 described tools plus a generic one for the whole API), Monta (remote at https://partner-api-mcp.monta.app/mcp, one tool per endpoint) and AMPECO (the @ampeco/public-api-mcp package, which you run on your own machine with npx).

Which platform ships a ready-made AI agent?

AMPECO, with CoOperator: a built-in conversational agent that diagnoses issues and runs actions after you confirm them. It has a free tier and paid tiers. askacharge.com and Monta provide the infrastructure (API, MCP, permissions, webhooks) for you to bring your own agent.

How does an agent know an OCPP command actually worked?

On askacharge.com every command returns a job_id that moves from 'acked' to 'confirmed' when the event that proves it arrives (for example charger.online after a Reset), or to 'timeout' if it doesn't. We did not find an equivalent mechanism in the public Monta and AMPECO documentation we reviewed.

EV charging management software for AI agents: askacharge.com, Monta and AMPECO compared (2026)

If an AI agent is going to run a charging network, "has an API" is no longer enough. What matters is what you let it do, how it knows an order worked, what happens when it repeats one, and where it can practise without breaking anything. We compared the three platforms that publish tooling for agents, using their own documentation as of 29 September 2026. We are one of the three, so every data point carries its source, and we also say where the other two do better.

Two different approaches

askacharge.com and Monta provide the infrastructure for you to bring your own agent: API, MCP server, permissions and events. AMPECO also has an API and an MCP server, but its main bet is CoOperator, an agent built into its platform. If you want to build your own agent (or sell one), the first two suit you better; if you want a ready-made agent and don't mind it being your vendor's, AMPECO is ahead.

The comparison

	askacharge.com	Monta	AMPECO
MCP server	Remote (`/api/mcp`), nothing to install. 25 hand-described tools plus a generic one reaching all 468 API operations.	Remote (`partner-api-mcp.monta.app/mcp`). One tool per API endpoint, generated automatically.	`@ampeco/public-api-mcp` package you run on your machine with npx (or host yourself). Generated from its OpenAPI.
Tools filtered by permissions	Yes: only what the key can use.	Yes: only what the credential can use; with read and write, delete tools are hidden.	Not found in its documentation.
Per-command OCPP permission	Yes: `commands:reset`, `commands:remote-start`… (10 commands), each separate.	Not found in its documentation: permissions are read / write / delete.	Not found in its documentation.
Limit to one location	Yes, for acting and for reading: `commands:reset@hotel-madrid`. Brand-wide aggregates answer 403.	Yes, credentials restricted to teams. Webhooks still arrive for the whole operator.	Not found for the API. CoOperator is isolated per operator.
Confirm a command took effect	Yes: `job_id` with acked → confirmed / timeout / rejected / failed, closed by the real event (e.g. `charger.online` after a Reset).	Not found in its documentation.	Not found in its documentation.
Retries without duplicates	Yes: `Idempotency-Key` header on any write.	Not found in its documentation.	Not found in its documentation.
Human approval before acting	Your agent decides; the platform limits it through permissions.	Monta's Claude plugin asks to confirm every write.	CoOperator previews every consequential action and waits for your confirmation.
Webhooks	8 events signed with HMAC-SHA256, with retries. Configured per brand.	Documented as not guaranteeing delivery or order; reconciliation recommended. Always operator-wide.	Yes (API and notifications).
Log of what the agent did	Writes and refused attempts for every key, 90 days.	Not found for the API.	Everything CoOperator does is logged.
Testing without hardware	Up to 10 simulated chargers inside your account that answer commands. No separate test environment.	Open-source OCPP emulator (desktop and web): 1.6 complete, 2.0.1 in development.	Not found in its documentation.
Built-in agent	Not offered to customers.	No, beyond the Claude plugin.	Yes: CoOperator (conversational). Background autonomous workers announced as "coming soon".
Published price	Yes: €20/month, unlimited chargers.	Yes: Starter €300 first year + 5 %; Advanced from €500/month + per outlet + 2 %.	No. CoOperator has a free tier and paid tiers on request.

"Not found in its documentation" means exactly that: it is not in the public documentation linked below. It may exist. If you have a source, write to info@askacharge.com and we will correct the table.

Where Monta does better

MCP coverage. One tool per API endpoint: the agent reaches everything without going through a generic tool.
An open-source OCPP emulator that also lets you intercept messages to reproduce misbehaving chargers. Our simulated chargers live inside your account and cannot do that.
Ecosystem: an official Claude plugin, a separate MCP for its docs and a well-established driver app.

Where AMPECO does better

A ready-made agent. CoOperator diagnoses failed sessions and authorisations, changes tariffs or creates chargers in plain language, with confirmation before every action. We offer nothing equivalent.
Stated isolation of each CoOperator instance per operator, and traceability of everything it does.

Where askacharge.com does better

Knowing whether the order worked. A Reset that returns "Accepted" doesn't tell you the charger came back. Our job_id confirms it with the real event; in our published test, after 21 seconds.
Permissions shaped for an agent: per command and per location, including what it can read. An agent can reset in Madrid and nothing else.
Safe retries with Idempotency-Key, and a log of every action and refused attempt.
Price: €20/month with unlimited chargers, MCP and API included.

See it working

We published the full test, in production and with the real responses: an AI agent that watches and resets chargers, with video. You can repeat it in your account with 5 simulated chargers.

Sources (checked on 29 September 2026)

Monta: Partner API MCP, Claude plugin, webhooks, OCPP emulator, pricing.
AMPECO: CoOperator, @ampeco/public-api-mcp, public API.
askacharge.com: technical documentation and the agent test.