The service's capabilities are handed to an AI assistant as tools over the MCP protocol. Accounts, audiences, mailings, statuses and replies are the same as in the API, but the task only has to be explained in words. Works with Claude, ChatGPT, YandexGPT, Cursor, n8n and any MCP client.
Claude
assistant by Anthropic
ChatGPT
assistant by OpenAI
Gemini
assistant by Google
YandexGPT
assistant by Yandex
Copilot
assistant by Microsoft
Cursor
code editor
n8n
automation
Work where you already work: you connect the assistant that you and your team already use, so there is no new interface to learn and no one to retrain — tasks are set in words right in its window. Nothing needs to be installed, and the list is open: if your assistant is not here, write to support — most likely it will work too.
The assistant sees the connected WhatsApp and Telegram accounts and the Telegram bots, sees their state and can connect a new one
Creating a draft with messages and recipients, starting, stopping and cancelling it — at your request, in plain words
Contact segments, adding numbers and field values — name, city, order amount — straight from the chat with the assistant or from a list file: Excel, CSV, a document or vCard
The pace is chosen by the autostrategy, not by the assistant: the service lowers the risk of an account ban and pauses the mailing itself if the messenger has restricted the account
A check of whether the numbers exist in the messenger: unregistered ones are filtered out and returned to the assistant with a separate status
Inbound messages, conversation history and replies to a mailing — the assistant reads them at your request, makes sense of them and answers the customer in the same dialog, quoting the message when needed
Sent, delivered and read statuses for every recipient and a summary for the mailing as a whole — with no exports and no reports
The plan and the remaining quota, the capabilities of each messenger and the rules of safe sending — the assistant reads them itself and does not suggest the impossible
Preparing a mailing and starting it are separate actions: the draft is visible in your account, and sending starts only once you have agreed
There is no password: sign in with Yandex, Google or email.
Open the API section in your mailing account and click “Create key”. The token is shown once — save it straight away.
The server address is https://mcp.wsender.io/. Paste it into the assistant's connection settings and put your token in the Authorization header. No program has to be installed — the server runs remotely.
WhatsApp or Telegram accounts — by QR code or phone number, bots — by token. In your account or through the assistant.
Ask the assistant to collect an audience, prepare a mailing, start it and show the result — in plain words.
AI integration pays off where working with messengers consists of many small actions and describing the task is easier than programming it.
A segment is described in words: “those who replied but did not buy”. The assistant finds them, prepares the text and shows the draft before sending.
The assistant reads customers' replies, separates those who are interested from those who refused and suggests who to write to today and with what offer.
Routine questions are closed without a manager: the price list, the terms and the instructions go out at once, and a difficult dialog is handed over to a person.
Invitations go out across the candidate base, and the assistant sorts the responses itself: who agreed, who refused and who to write to later.
A survey after a purchase or a visit, and instead of reading hundreds of conversations — a summary by meaning: what gets praised and what gets complained about most often.
Ask “how did yesterday's mailing go” and delivery, reads and replies come back as plain text, with no exports and no spreadsheets.
Topping up the base from a file or from a conversation, checking numbers in the messenger, keeping segments and contact properties — without opening your account.
Scenarios on n8n or your own platform: the agent starts touchpoints on a schedule and on events in your system, with no person involved.
One-off tasks on top of your CRM: a selection by deal stage or a mailing “to one country only”.
The set of capabilities is the same as in the messenger API: the MCP server works on top of it and on the same keys, so automation in your systems and the assistant's work live in one project and see the same accounts, segments, mailings and customer replies. The difference is the way in: the API describes a scenario in code once and for all, while the assistant only needs the task explained in words — and changed at any moment.
MCP (Model Context Protocol) is an open standard by which AI assistants connect to external services. Put simply, it is a socket: the service declares which actions it can perform, and the assistant performs them at the user's request. The standard is supported by Claude, ChatGPT, YandexGPT, Gemini, Copilot, Cursor and automation platforms such as n8n — you connect once and it works in every scenario of that assistant.
Any that can connect remote MCP servers: Claude, ChatGPT, YandexGPT, Cursor, code editors with MCP support, automation platforms such as n8n and your own agents built on models. No separate program has to be installed — the server works at an address on the internet, and the assistant's settings only need that address and your key.
If you are not sure whether your assistant supports MCP, write to support and we will check it together.
The capabilities are the same — this is one integration served two ways. The API is for your system: a developer describes the scenario once, and after that messages go out automatically from a CRM, a shop or a website. Every new rule, though, has to be programmed.
MCP is for a person and their assistant: the scenario is described in words and changes in a minute. One-off and non-standard tasks — “do the same, but for one country only” — need neither development nor waiting for a contractor. One does not get in the way of the other: both work in the same project and on the same keys.
There is no separate fee for connecting — the usual plans of the service apply, just as for the API. After your first sign-in the free Basic plan is enabled with 15 messages a day, forever — enough to try working with an assistant.
For working volumes there are paid plans — their limits and prices are on the plans page.
One caveat about the free plan: mailings are created and sent on it, but reading recipients' replies, dialogs and account contacts requires a plan with account synchronisation. A subscription to the AI assistant itself is paid to its vendor separately.
No. On its own the assistant does nothing: it receives no notifications about events and does not watch the correspondence — every action starts with your request in the chat. Preparing a mailing and starting it are separated as well: the assistant creates a draft, shows a link to it in your account and starts sending only after your confirmation.
Automatic replies to inbound messages are a separate feature of the service: a greeting for a new bot subscriber and a reply to a message are set up by channel rules in your account, work without the assistant and can be switched off at any moment.
Messages go out by the usual rules of the service: from your accounts, within the volume the plan allows and at a safe sending pace — the limits cannot be bypassed through the assistant. Access is granted by the API key, and deleting it in your account disconnects the assistant immediately.
No. Unlike an API integration, nothing is programmed here: you paste the server address and your key into the assistant's settings once, and after that tasks are set in plain text. Create an API key in your account, paste the address https://mcp.wsender.io/ and that key into the assistant's settings — and that is all. If something does not work out, write to support: we will help with the setup — and do tell us which tasks you would like to hand over to the assistant, it affects what we build next.
