MCP Overview
Wooxy MCP Server
Connect any MCP-compatible AI agent to your Wooxy account. The agent reads your account structure, builds and previews branded templates, defines audience segments, schedules campaigns, and pulls sending statistics — through the same permissions your user already has in Wooxy.
Wooxy runs a remote MCP server. There is nothing to install and no local process to keep alive: add the endpoint as a connector, authenticate, and the tools appear in your agent.
Quick start
1. Open your agent's connector settings
In an MCP-compatible client (Claude, Cursor, VS Code, Gemini CLI, Windsurf, Goose, and others) open Connectors, MCP servers, or Integrations, and choose Add custom connector.
Name the connector Wooxy and enter the remote server URL:
https://mcp.wooxy.com/mcp
The server speaks MCP over Streamable HTTP. Clients that only support the older HTTP+SSE transport should use their compatibility mode.
3. Authorize the connector
Click Connect. Wooxy takes over the flow from there:
You sign in to your Wooxy account, as you would on the website
A consent screen lists the permissions the connector is asking for.
You confirm, and Wooxy redirects you back to your agent.
There is no API key to generate, copy or rotate. The connector is bound to the Wooxy user who authorized it, and every tool call runs as that user, inside that account, with that user's permissions — so a connector cannot reach anything the person behind it could not already open in the Wooxy UI.
That last point is why several tools return partial responses rather than errors. context__get, for instance, includes billing details only for a user who holds the billing permission, and the domain list only for a user who holds the settings permission. If a block you expected is missing, check the permissions granted at this step before assuming the account has no data.
4. Start working
Ask your agent something concrete, for example:
The agent calls the Wooxy tools directly. It reads your domains, lists, templates and reports, and asks you before it schedules anything.
Menu names vary slightly by agent. Check out per-product setup guides in the side menu.
Configuration snippet
Clients that are configured from a file (rather than a UI) use a shape like this:
Tool reference
Session
| Tool | Description |
|---|---|
| context__get | Current user, account, timezone, balance and plan, plus embedded brands, contact sources and custom events. |
Account structure
| Tool | Description |
|---|---|
| domains__list | Sending domains and their verification state. |
| contact_lists__list | Contact lists (audiences) belonging to one sending domain, with their personalization variables. |
| contact_sources__list | Acquisition sources and their customer acquisition cost. |
| custom_events__list | Registered custom events available for behaviour segmentation. |
| brands__list | Brands available for template generation. |
| telegram_bots__list | Telegram bots, optionally filtered to one domain. |
Audience segmentation
| Tool | Description |
|---|---|
| segments__list | Segments defined on one contact list. |
| segment__create | Create a reusable saved filter over a contact list. |
| segment__update | Rewrite an existing segment's filter definition. |
Templates
| Tool | Description |
|---|---|
| templates__list | Template metadata across channels — id, name, channel, creation time. Never the body. |
| template__create | Generate a branded template from a free-form idea. Returns a preview. |
| template__modify | Produce a new revision of a generated template from a free-form change request. |
| template__save | Persist a generated template into the account so campaigns can use it. |
Campaigns
| Tool | Description |
|---|---|
| campaigns__list | Campaigns in the account, newest first. |
| campaign__schedule_create | Create a one-time campaign that launches once at a future date and time. |
| campaign__repeat_create | Create a recurring campaign on a daily, weekly or monthly cadence. |
Reporting
| Tool | Description |
|---|---|
| reports__stat | Raw account-level sending metrics for a date range, optionally scoped by channel and domain. |
| campaign__report | Per-iteration counters and rates for one campaign, with geo and device breakdowns. |
| template__report | Counters and rates for one template over a date range, bucketed by hour, day or month. |
| contact_list__report | Composition of one audience — size, per-channel reachability, blacklists, demographics. |
Documentation
| Tool | Description |
|---|---|
| tool_docs__get | Full reference documentation for any tool on this server. |
| kb__tree | Knowledge base categories and article ids. |
| kb__article | Full text of one knowledge base article. |
| api_docs__overview | Index of the Wooxy HTTP API, by section and method id. |
| api_docs__method | Full documentation for one HTTP API method, including request and response examples. |
FAQ
Do I need an API key?
No. You sign in to Wooxy, approve the permissions on the consent screen, and get redirected back to your agent. Every call then runs as your user, with your permissions.
Which permissions does the connector get?
The ones shown on the consent screen at authorization time, scoped to the account you signed in to. Nothing is granted silently, and the connector never exceeds what your own Wooxy user can do.
Is there a local server to install?
No. Wooxy MCP is a remote server. Add the endpoint URL and authenticate.
Can the agent send a campaign without asking me?
It can create one, which schedules a real send. The server marks campaign__schedule_create and campaign__repeat_create as write tools that need confirmation, and well-behaved clients prompt for approval before every call. Review the campaign in Wooxy before its launch time if you want a second check.
Can the agent read my template HTML?
No. templates__list returns metadata only — id, name, channel and creation time. Generated templates are shared as a PNG preview and a link.
How does the agent learn the filter grammar?
It asks the server. tool_docs__get("segment__create") returns the complete rule matrix at runtime, which is why an agent can build a correct segment without that grammar being in its training data.
What about the HTTP API?
It is still there, and the MCP server can document it: api_docs__overview lists every section and method, and api_docs__method returns one method's full reference with code examples. MCP wraps Wooxy 's capabilities so an agent can use them conversationally; the HTTP API is what you build against directly.
