Systeme.io Extra MCP

Systeme.io MCP Server: Use AI to Automate Sales Funnels, Email Marketing and CRM

Run your Systeme.io account from your AI assistant: contacts, tags, funnels, pages, emails, courses, products, payments and more.

Systeme.io Extra MCP covers every endpoint in Systeme.io's public API documentation: 152 actions in 22 tool groups. It works on its own, without the official Systeme.io MCP, and connects to assistants such as Claude, Codex, Antigravity and Cursor.

🚀 Don't have an account yet? Create your free Systeme.io account to get started with the platform. Already have one? Use your existing account.

Affiliate link: SoftReviewed may earn a commission from eligible purchases.

Here, automation means asking a connected assistant to perform supported actions; it does not mean an unattended business that runs without your instructions or review.

Built by SoftReviewed. Independent project; not endorsed by Systeme.io.

Quick start: let your AI install it

1. Create a Public API key in Systeme.io: Settings -> MCP & API keys -> Public API keys -> Create. (Not the MCP key on the same page; that one only works for the official MCP.)
2. Paste this into your AI assistant (Claude Code, Codex, Cursor, Antigravity and similar tools that can run commands):

Install https://github.com/softreviewed/systeme-extra-mcp for me. Follow its AGENTS.md. Keep my other MCP servers. Do not ask me to paste my API key in chat.

3. Save your key when the assistant asks (a hidden prompt, or one line in your config file), then restart the app.
4. Ask: "List my Systeme.io tags." If you see your tags, you are ready.

The assistant picks the setup that fits your computer:

Your computer has	Setup used	Where the key is kept
Docker Desktop with MCP Toolkit	Docker Gateway	Docker's secret store (safest)
Python 3.11 or newer	Direct Python	Your assistant's config file

Prefer to do it yourself? Follow Connect your assistant.

Guide

• Quick start
• What you can do
• Official Systeme.io MCP or this one?
• Is it right for you?
• Connect Systeme.io to Claude, Codex, Antigravity or Cursor
• Technical reference
• Troubleshooting
• Verification limits

Why use it?

You want to tag a customer, check last week's newsletter results, give a student course access and pause a coupon. Normally that is four dashboard screens.

With this server your assistant can do all of it in one conversation. Every change is previewed first, and nothing is written until you say yes.

Your request -> assistant finds the right record -> previews the change -> you authorize it -> account updated.

An assistant still needs accurate instructions and your review. The connection does not replace your judgement or guarantee that an assistant selects the right customer.

What you can do

Area	Tool groups	Examples
Contacts and CRM	contacts, tags, contact_fields	Find a contact by email, add or remove tags, define custom fields
Automation	automation_rules, automation_emails	List, create and edit rules; manage automation emails
Funnels and pages	funnels, pages	List funnels and steps, create a step, read or save page content, apply a template
Email marketing	newsletters, campaigns, email_statistics	Draft, target, schedule, test and send newsletters; edit campaign steps; read open and click stats
Courses and community	courses, enrolments, communities	Build courses, modules and lectures; grant or remove access; manage community members
Selling	products, price_plans, coupons, sale_settings, subscriptions	Manage digital and physical products, offers, prices and coupons; cancel a customer subscription
Scheduling	booking	Read bookings; manage calendar events and availability
Integrations	webhooks, sms	Manage event webhooks; check Twilio setup and SMS templates
Account	account	List connected domains and uploaded files

Every action, its method, path and required inputs are listed in the feature reference. Ask the assistant for help on any group to see the same list.

Example requests

• "Find the contact jane@example.com and preview adding the tag Customer."
• "Show open and click rates for my last three newsletters."
• "Preview a 20% coupon for my course, valid until the end of the month."
• "Show who has access to this course, then preview giving this customer access."
• "Find this customer's subscription and preview cancelling it at the end of the billing period."

Safety built in

Type of action	What happens
Read	Runs straight away
Write	Blocked until previewed (dry_run) and confirmed (confirm) after you agree
High-impact write	Same gate, flagged as high_impact in help: sending a newsletter, deleting contacts, courses, campaigns or products, cancelling a subscription, overwriting a page

Only documented endpoints are allowed. Two deprecated endpoints (PUT newsletter and coupon updates) are left out in favour of their PATCH versions.

Official Systeme.io MCP or this one?

	Official Systeme.io MCP	Systeme.io Extra MCP
Coverage	Its own selected tool set	Every endpoint in the public API docs (152 actions)
Course enrolments, community members, custom field definitions, subscriptions, webhooks, Twilio checks	Not in the inventory we inspected	Included
Hosting	Hosted by Systeme.io	Runs on your computer (Docker Gateway or direct Python)
Credential	Official MCP key	Public API key (Docker's secret store or your config file)
Write protection	Its own rules	Preview, then explicit confirm for every write

You can use either one alone or both together. See the official Systeme.io MCP overview for what it offers. The official inventory was inspected on 6 October 2026; it may change.

Which key? The official MCP uses an MCP key. This server uses a Public API key. Both are under Settings -> MCP & API keys, in separate lists.

Is it right for you?

Advantages

• The whole public API in one connection: 22 tool groups across contacts, email, funnels, courses and payments.
• Works without the official MCP: useful when you want one local connection for every assistant.
• Preview before changing account data: every write is previewed and needs your go-ahead.
• Two ways to install: Docker Gateway keeps the key in Docker's secret store; direct Python needs no Docker at all.
• Portable setup: rebuild the image and recreate your secret when moving computers.

Trade-offs and limitations

• Setup takes a few minutes. You need Docker Desktop or Python 3.11+, and a Systeme.io Public API key. An assistant that can run commands can do the setup for you.
• Not every action is tested live. All 152 are checked offline; live checks cover read actions only. See verification limits.
• Only what the API offers. Anything missing from Systeme.io's public API (for example, refunds or sending a single SMS) is missing here too.
• Your account's limits still apply. It does not unlock paid features, raise capacity or bypass permissions.
• No built-in cloud hosting. It runs on your computer; a cloud-only assistant needs a separately hosted, authenticated connection.

New to Systeme.io?

You need a Systeme.io account before this server can work with your data. Create a free Systeme.io account, then learn the dashboard before connecting an assistant to customer records.

For a small starting project, read our Systeme.io free-plan features and limits guide. If you later need more capacity, compare the Systeme.io pricing plans and our explanation of annual discounts and savings. A higher plan should solve a real capacity need; installing this MCP does not itself require you to upgrade.

If your goal is to build funnels for yourself or clients, our Systeme.io Funnel Builder certification guide covers that learning path. You can also compare the two Systeme.io certification courses before choosing where to start.

You can also visit SoftReviewed for additional resources.

Affiliate disclosure: SoftReviewed may earn a commission from eligible purchases made through its referral links. This does not make the MCP an official Systeme.io product.

Let your assistant handle the setup details

Share this repository URL with your assistant and use the prompt from the Quick start. Step-by-step instructions written for assistants are in AGENTS.md: how to choose Docker or direct Python, where each app keeps its config, how to store the key safely and how to check it works.

An assistant can read the instructions, but it needs to be able to run commands on your computer to set it up (for example Claude Code, Codex, Cursor or Antigravity). A chat-only assistant can still walk you through the steps. Never put your API key into a public issue or commit.

The remaining sections are for installation and technical reference. Exact action names, required inputs and edge cases are in the detailed feature reference; copyable requests are in the operation examples.

Connect your assistant

Two ways to run it. Both give the same 22 tools.

	Option A: Docker Gateway	Option B: Direct Python
You need	Docker Desktop with MCP Toolkit	Python 3.11 or newer
Key stored in	Docker's secret store	Your assistant's config file
Best for	Keeping keys out of config files; sharing one Gateway	The quickest setup, no Docker

Option A: Assistant -> Docker Gateway -> container -> Systeme.io API. Option B: Assistant -> Python on your computer -> Systeme.io API.

Option A: Docker Gateway

Install/start Docker Desktop with MCP Toolkit/Gateway available. From this repository, check the CLI and build the image:

docker version
docker mcp gateway run --help
docker build -t mcp-systeme-extra:v2 .

The image includes its source and schemas; clients need no source-folder mounts or local Python installation. Copy docker-catalog.yaml to ~/.docker/mcp/catalogs/systeme-extra.yaml (Windows: %USERPROFILE%\.docker\mcp\catalogs\systeme-extra.yaml), creating the folder if needed.

New-Item -ItemType Directory -Force "$env:USERPROFILE\.docker\mcp\catalogs" | Out-Null
Copy-Item -LiteralPath .\docker-catalog.yaml -Destination "$env:USERPROFILE\.docker\mcp\catalogs\systeme-extra.yaml"

Check discovery:

docker mcp gateway run --catalog systeme-extra.yaml --servers systeme-extra --dry-run

Expected: 22 tools, named systeme_extra_<group>. Explicit --servers selects this server without relying on the enabled-server registry. For an existing shared Gateway, merge the catalog entry, preserve other servers and reuse the existing client connection instead of adding duplicates.

The catalog workflow was tested with the installed CLI. Newer Docker Toolkit profile workflows may differ; inspect your installed docker mcp gateway run --help. Docker MCP documentation.

Credentials

🚀 New to Systeme.io? Create your free account here, then follow the API-key steps below. Already registered? Go straight to your settings.

Affiliate link: SoftReviewed may earn a commission from eligible purchases.

Before using account actions, create your own public API key:

1. Sign in and open Systeme.io API-key settings, or use your profile picture -> Settings -> MCP & API keys.
2. In Public API keys, select Create. Give it a recognizable name such as Systeme.io Extra MCP, choose an expiration appropriate for your setup, then save.
3. Store the key securely using the connection method below. The official MCP credential is a different key and cannot replace this public API credential.
4. Start with help, a preview and a harmless account read before requesting changes. Help and previews can run without the key; account reads and writes need it.

See the official API-key guide if the dashboard labels change. Never post the key in a GitHub issue, screenshot or public file. Remove/revoke it in your account when you no longer use this connection.

For Docker Gateway, store it as Docker secret systeme.api_key, injected as SYSTEME_API_KEY.

PowerShell input avoids placing the key in command history:

$systemeSecureKey = Read-Host "Systeme.io public API key" -AsSecureString
$systemeKeyPointer = [Runtime.InteropServices.Marshal]::SecureStringToBSTR($systemeSecureKey)
try {
    [Runtime.InteropServices.Marshal]::PtrToStringBSTR($systemeKeyPointer) | docker mcp secret set systeme.api_key
} finally {
    [Runtime.InteropServices.Marshal]::ZeroFreeBSTR($systemeKeyPointer)
    Remove-Variable systemeSecureKey, systemeKeyPointer
}

With Option A, no credentials belong in client JSON. With Option B, the key sits in your assistant's config file on your computer only; never commit or share that file. Help and previews work without credentials; account requests do not.

Option B: Direct Python (no Docker)

Download the code and install its requirements once:

git clone https://github.com/softreviewed/systeme-extra-mcp.git
cd systeme-extra-mcp
python -m venv .venv

Then, on Windows: .venv\Scripts\python -m pip install -r requirements-lock.txt. On macOS/Linux: .venv/bin/python -m pip install -r requirements-lock.txt.

Add this entry to your assistant's config, using the full path to your folder. On Windows the Python path ends in .venv/Scripts/python.exe.

{
  "mcpServers": {
    "systeme-extra": {
      "command": "/ABSOLUTE/PATH/systeme-extra-mcp/.venv/bin/python",
      "args": ["/ABSOLUTE/PATH/systeme-extra-mcp/server.py"],
      "env": { "SYSTEME_API_KEY": "YOUR_PUBLIC_API_KEY" }
    }
  }
}

Replace YOUR_PUBLIC_API_KEY with your key, save, then fully restart the app. Ready-made files for each app are in examples/clients/direct: Claude Desktop, Cursor, Antigravity (JSON) and Codex (TOML). For Claude Code:

claude mcp add --transport stdio --scope user systeme-extra --env SYSTEME_API_KEY=YOUR_PUBLIC_API_KEY -- /ABSOLUTE/PATH/systeme-extra-mcp/.venv/bin/python /ABSOLUTE/PATH/systeme-extra-mcp/server.py

The sections below show the Option A (Docker) entry for each app. For Option B, use the same file locations with the entry above.

Several Systeme.io accounts in one server (Option B)

Manage a shop, a blog and client accounts with one install. You choose each account's name; your assistant switches when you ask.

1. Create a Public API key in each account. Systeme.io → Settings → MCP & API keys → Public API keys. Use the Public API key, not the MCP key.

2. Name your accounts in the config. In the env block of your server entry, add one line per account: SYSTEME_API_KEY_ followed by the name you want.

"env": {
  "SYSTEME_API_KEY_SHOP": "PUBLIC_API_KEY_FOR_SHOP",
  "SYSTEME_API_KEY_CLIENT-2": "PUBLIC_API_KEY_FOR_CLIENT_2"
}

You write in the config	You say in chat
SYSTEME_API_KEY_SHOP	"shop"
SYSTEME_API_KEY_CLIENT-2	"client-2"
SYSTEME_API_KEY_MY_BLOG	"my_blog"

Naming rules: letters, digits, _ and - only (no spaces; write my_blog, not my blog). Names are not case-sensitive. default and all are reserved. Keep your keys in the config file; never paste them into a chat. For Claude Code, repeat --env once per account: claude mcp add ... --env SYSTEME_API_KEY_SHOP=... --env SYSTEME_API_KEY_CLIENT-2=....

3. Restart your app once, then say which account you mean:

• "Show the contacts in client-2."
• "List the tags in shop, then the tags in blog."
• "Compare my courses across all accounts."

The assistant passes account on each call, so switching needs no config edit and no restart. To rename or add an account later, edit the config and restart once.

Situation	What happens
Only SYSTEME_API_KEY is set (your current setup)	Works exactly as before; no account name needed
account: "shop"	Uses that account's key
account: "all"	Reads from every account; each result is labelled with its account
Several accounts, no name given	Reads use the account called default (from SYSTEME_API_KEY) if you keep one; otherwise the assistant asks which account
A write (create, update, remove, send)	Always needs one named account; all is refused. Previews (dry_run) show the account that will be used
Unknown account name	Rejected, and the reply lists the valid names

Every reply includes the account it came from, and help lists the account names, never the keys. Option A (Docker) stores one secret, so it stays single-account; use Option B for several accounts.

Claude Desktop

To connect Systeme.io to Claude Desktop using this server, configure the local Docker Gateway entry below. This is separate from the official hosted Systeme.io connector.

Settings -> Developer -> Edit Config. Merge claude-desktop.json, then fully restart the app.

Windows config: %APPDATA%\Claude\claude_desktop_config.json. macOS: ~/Library/Application Support/Claude/claude_desktop_config.json. Official local-server guide.

{
  "mcpServers": {
    "systeme-extra": {
      "command": "docker",
      "args": ["mcp", "gateway", "run", "--catalog", "systeme-extra.yaml", "--servers", "systeme-extra"]
    }
  }
}

Claude Code

claude mcp add --transport stdio --scope user systeme-extra -- docker mcp gateway run --catalog systeme-extra.yaml --servers systeme-extra
claude mcp list

Use /mcp to inspect status. Project-only configurations belong in .mcp.json rather than a user-level connection. Claude Code documentation.

Codex

codex mcp add systeme-extra -- docker mcp gateway run --catalog systeme-extra.yaml --servers systeme-extra
codex mcp list

Alternatively merge codex.toml into ~/.codex/config.toml:

[mcp_servers.systeme_extra]
command = "docker"
args = ["mcp", "gateway", "run", "--catalog", "systeme-extra.yaml", "--servers", "systeme-extra"]
startup_timeout_sec = 60

Reload the local app/extension; /mcp shows CLI status. Hosted ChatGPT web does not read this local configuration. Codex MCP documentation.

Google Antigravity

MCP Servers -> Manage MCP Servers -> View raw config. Merge antigravity.json, using the same mcpServers entry as Claude Desktop.

Current documented paths: global ~/.gemini/config/mcp_config.json, project .agents/mcp_config.json. Reload/check tools. Paths and menus vary by release; use the file opened by your installed app. Antigravity documentation.

Cursor

Merge cursor.json into global ~/.cursor/mcp.json or project .cursor/mcp.json, then reload/check MCP status.

{
  "mcpServers": {
    "systeme-extra": {
      "type": "stdio",
      "command": "docker",
      "args": ["mcp", "gateway", "run", "--catalog", "systeme-extra.yaml", "--servers", "systeme-extra"]
    }
  }
}

Cursor documentation.

Other local clients / cloud hosting

For another stdio client, use command docker with the same arguments. If Docker cannot be found, use its absolute executable path from Get-Command docker (Windows) or command -v docker (macOS/Linux). Merge configurations without replacing other servers.

This repository does not provide a hosted MCP URL. Cloud-only assistants cannot start your computer's Docker process. Remote access requires a separately deployed, authenticated Gateway/transport and compatible client; local JSON is not a cloud deployment recipe. Do not expose an unauthenticated Gateway publicly.

Technical reference

The operations are generated from Systeme.io's public API documentation. Use it to verify supported endpoints alongside the checked-in schemas and examples.

Common arguments

Argument	Meaning	Default
action	Operation, or help for exact schemas	help
parameters	Path IDs and query filters	Empty
body	Payload for operations that accept one	Empty
dry_run	Validate/preview without requesting the API	false
confirm	Permit a write after human authorization	false
account	Which configured Systeme.io account to use; reads also accept all	Only account, or default for reads

Workflow: help -> find real IDs -> preview -> user authorizes -> execute -> verify. confirm:true is a server guard, not evidence of human permission. All write examples in examples/operations.json are previews; execute an authorized write by replacing dry_run:true with confirm:true. Replace example IDs with actual selected records.

Implementation

Detail	Behavior
Runtime	Python 3.11 image; FastMCP 2.14.2
Dependencies	Docker installs pinned requirements-lock.txt
Transport	Server stdio through Gateway
Authentication	https://api.systeme.io; X-API-Key header
HTTP	httpx; 30-second timeout; redirects disabled
Validation	Official schemas; unknown fields rejected; path IDs URL-encoded
Writes	POST/PATCH/DELETE guarded; PATCH uses application/merge-patch+json
Retries	No automatic retries, including writes
Output	API data preserved; sensitive field names redacted recursively

List actions that support paging take limit (10-100), order (asc/desc) and startingAfter. When the result reports hasMore, pass the last item's ID as the next startingAfter, keeping the same filters and order. There is no automatic all-page retrieval. Use help with parameters={"action": "<name>"} to see the exact inputs of any action.

Success includes ok, status, data, rate_limit_remaining. HTTP errors include ok:false, status, documented explanation, and retry_after when returned. No fixed API allowance is promised. A timeout does not prove a write failed: inspect account state before retrying.

Redaction is field-name based, not complete removal of personal data. Do not share customer outputs publicly. All 152 method/path pairs and original schemas/source URLs are in docs/operations.json; copyable inputs are in examples/operations.json.

Endpoint map

The full map of 152 actions, with method, path, required input and read/write type, is in the feature reference. It is generated from docs/operations.json by scripts/build_reference.py; docs/operations.json itself is rebuilt from the official docs by scripts/build_operations.py.

Troubleshooting

Symptom	Next step
Docker not found	Start Desktop; use absolute executable path
No tools	Check image tag, catalog path and server name
Duplicate tools	Reuse the existing Gateway connection
configured:false	Store Docker secret; missing/<UNKNOWN> keys cannot authenticate
HTTP 401 "Invalid API Key."	Use a key from Public API keys, not the MCP key; check it was not deleted
Invalid input	Use help; verify spelling, types and required fields
Partial access rejected	Supply a non-empty integer modules array
Subscription list cannot run	Find a real contact ID first
Write blocked	Preview with dry_run, get approval, then send with confirm
SMS 424 / account 404	Connect Twilio; documented missing-integration responses
HTTP 429	Respect returned retry/rate-limit information
Write timed out	Check account state before retrying
Source edits invisible	Rebuild image and restart client/Gateway
ModuleNotFoundError (Option B)	The config must point at the .venv Python, not the system one
Tools missing after config edit	Fully quit and reopen the app

Verification limits

Check	Status, 13 October 2026
Local tests	20 passed; previews for all 152 actions, write gates, help and mocked HTTP
Docker Gateway	22 tools and 152 operations listed; write preview passed; no API requests sent
Direct Python (no Docker)	Clean copy installed from requirements-lock.txt on Python 3.11; client connected over stdio, 22 tools, write preview passed
Authenticated reads	28 of 31 read-only checks returned 200, including a subscription list; 3 Twilio checks gave the documented "not configured" replies (424 / 424 / 404)
Live writes	Not performed; local validation only
Individual assistant apps	Configs checked against documentation; not all tested in-app

Earlier v1 checks (6 October 2026): enrolment, community, membership and webhook lists returned 200, and missing-Twilio responses 424 / 424 / 404 matched the docs.

See full verification details. Account permissions and product restrictions still apply.

Development and portability

Create a Python 3.11 virtual environment:

python -m venv .venv

Activate with .venv\Scripts\Activate.ps1 on Windows or source .venv/bin/activate on macOS/Linux, then run:

python -m pip install -r requirements-lock.txt
python -m unittest -v
python tests/gateway_smoke.py

The smoke test uses this repository catalog by default and sends no account API requests. Override its catalog with SYSTEME_EXTRA_CATALOG if needed. The separate live-read script requires credentials.

Moving computers: repository -> build image -> install catalog -> recreate Docker secret securely -> merge client config -> check help and a harmless read. Copying the repository does not transfer credentials. No developer-specific source mount is needed.