This repository is the public home and setup guide for the datavessel MCP server. datavessel is a hosted, remote MCP server — there's nothing to install or run. Point your AI client at the URL below with your own API key.

What it is

datavessel gives AI agents (Claude, Cursor, and any MCP client) read and write access to your entire marketing and ecommerce stack through a single connection. One agent, your whole stack — pull data and take action.

• Bring your own key (BYOK). Your data, your credentials.
• Hosted & remote. No local install; Streamable HTTP.
• 100+ typed tools across 10 connectors, plus 20+ guided workflows.

Connectors

Connector	What your agent can do
Google Analytics 4	Traffic, realtime, custom reports, property config
Google Search Console	Search performance, top queries/pages, sitemaps, keyword opportunities
Google Ads	Campaign, ad-group, keyword & search-term performance
Meta Ads	Campaigns, ad sets, ads, account & campaign insights
Shopify	Orders, products, customers, inventory, discounts — incl. writes
WooCommerce	Orders, products, customers, sales reports, coupons
Shopware	Catalog, orders, customers — create/update CMS & landing pages, SEO URLs, products
Slack	Channels, history, post messages & DMs
LinkedIn	Publish text & article posts
datavessel	Reports, alerts, domain checks & utilities

Example: "Find Search Console keywords where we rank well but get few clicks, rewrite those product titles in Shopify, and post the summary to #seo in Slack."

Setup

1. Create your account — sign up at www.datavessel.io and connect the data sources you want (Google, Shopify, etc.) via OAuth in the dashboard. See datavessel.io/mcp for the full MCP overview.
2. Add the server to your MCP client using the remote Streamable HTTP URL https://mcp.datavessel.io/mcp.

Authentication is handled automatically over OAuth — there is no API key to copy or paste. datavessel implements the MCP authorization spec: the first time you connect, your client discovers the server's OAuth metadata, registers itself dynamically (DCR), and opens a browser to sign you in. Tokens (with PKCE) are issued and refreshed by the client for you.

Clients with native remote MCP support (Claude, etc.)

Add the remote server by URL — your client runs the OAuth flow on first connect:

{
  "mcpServers": {
    "datavessel": {
      "type": "streamable-http",
      "url": "https://mcp.datavessel.io/mcp"
    }
  }
}

Clients without native remote support (via mcp-remote)

mcp-remote bridges a stdio client to the remote server and performs the same OAuth/DCR handshake — again, no token in the config:

{
  "mcpServers": {
    "datavessel": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.datavessel.io/mcp"]
    }
  }
}

On first connect a browser window opens to authorize datavessel; approve it once and the tools appear. If a connection ever fails with an auth error, clear your client's saved tokens and reconnect — it will re-register automatically.

Pricing

Tier	API calls	Price
Free	100 / month	$0
Professional	1,000 / month	$49 / month

All tools are available on both tiers — discovery calls don't count toward the limit.

Links & support

• 🌐 Website: https://www.datavessel.io
• 🚀 App / dashboard: https://app.datavessel.io
• 🔌 MCP details: https://datavessel.io/mcp
• ✍️ Blog: https://blog.datavessel.io
• 💬 Support: open an issue or contact us via the website

License

Apache-2.0 — see LICENSE.