VAT
Your cart

Your cart is empty.

MCP setup guides

These MCP setup guides explain how to configure each tool add-on for mcp-core, the self-hosted MCP server: which credentials you need, which env vars to set, which mode to pick and how to check that Claude sees the tools. Every add-on follows the same five steps; only the credentials differ.

Which MCP setup guides are there?

There is one guide per add-on. Pick the system you want Claude to work with:

  • Missive MCP setup: shared inbox. An API token; read-write adds sending, tasks and contact changes.
  • WooCommerce MCP setup: orders, products and customers through WooCommerce REST API keys. Several stores supported.
  • Database MCP setup: MySQL, MariaDB-compatible or PostgreSQL, with a SELECT-only user as the recommended setup.
  • Brain MCP setup: a markdown knowledge base in a folder on your server, with optional git versioning.
  • WordPress MCP setup: posts, pages, media, users, menus and settings through an application password. Several sites supported.
  • Lightspeed MCP setup: Lightspeed eCom C-Series data through an API key and secret. Always read-only. Several shops supported.
  • Plesk MCP setup: domains, DNS, databases and logs on your Plesk server through a Plesk API key. Read-only by default; read-write adds changes and commands.

How do the add-ons in the MCP setup guides compare?

The table below sums up what each guide covers: the entitlement on your license, the credentials, the default mode and whether one installation can talk to several instances.

Add-onEntitlementCredentialsDefault modeSeveral instances
Missivetool:missiveAPI tokenread-onlyNo
WooCommercetool:woocommerceConsumer key and secretread-onlyYes, WOOCOMMERCE_CONNECTIONS
Databasetool:databaseHost, name, user, passwordread-onlyNo
Braintool:brainNone, a vault folderread-writeNo
WordPresstool:wordpressUsername and application passwordread-onlyYes, WORDPRESS_CONNECTIONS
Lightspeed eComtool:lightspeedAPI key and secretread-only, no write toolsYes, LIGHTSPEED_CONNECTIONS
Plesktool:pleskPlesk API keyread-onlyNo

Which steps do all MCP setup guides share?

Every add-on is set up in the same order. If mcp-core itself is not running yet, start with the mcp-core install guide.

  1. Make sure the license holds the add-on’s entitlement. Without it, the tool is not exposed at all, whatever is in .env.
  2. Get the credentials from the system you connect, as described in each guide.
  3. Fill in the env vars in .env, in the folder you unzipped. The download already contains the required lines, empty, and the optional lines commented out.
  4. Make sure a seat in My Account → MCP Self-hosted → Seats has the tool ticked, with the right mode.
  5. Restart mcp-core with ./mcp-update.sh and check the tools from Claude.

Instead of .env, you can also enter the credentials in My Account → MCP Self-hosted → Tools. They are stored encrypted, a running mcp-core picks them up without a new download, and they override the same keys in .env.

What do the MCP setup guides mean for an add-on you buy later?

You don’t reinstall anything and you don’t download a new bundle. The add-on is added to your existing license, and you receive an email that lists the env vars the tool needs. Fill those in, then restart and refresh the license:

./mcp-update.sh
docker exec mcp-core npm run refresh-license

Without the manual refresh, mcp-core picks up the new entitlement at its next periodic license check. If AddoCore can’t be reached during a check, a 96-hour grace period keeps your current tools running; see license grace period.

How do tool modes work across the MCP setup guides?

Each tool runs read-only or read-write, and every guide lists exactly which tools each mode exposes. Read-only is the default for every tool except Brain, which defaults to read-write. You pick the mode per tool when you add a seat, so a teammate can get WooCommerce read-only while your own seat has read-write.

Each tool also has a <TOOL>_MODE env var, such as WOOCOMMERCE_MODE. The mode set on the seat takes precedence over it. Buying a tool and choosing its mode are separate controls; entitlements vs. tool modes explains why. How seats, tokens and per-seat tools work is covered in MCP seats.

How do you check a tool after following the MCP setup guides?

Ask Claude to call mcp_status. This built-in tool is always available, also during a license problem, and shows the mcp-core version, the license state, the active seat, its entitlements and the mode per tool. From the server itself, the health endpoint lists the tools that are active on the license:

curl http://127.0.0.1:3200/health

If a tool is entitled but missing in Claude, its credentials are usually incomplete. The container log then names the missing variables:

docker logs mcp-core

Each of the MCP setup guides ends with the error messages that tool produces and what to do about them. Start with the guide for the system you want to connect first.

Ready to connect your AI to real tools?

Start free trial