Use Erlin from an MCP client

A three-step diagram shows a user asking an AI client a question, the client retrieving Erlin data and the client explaining the returned results. — Your AI client uses Erlin data to answer your question.

The Erlin MCP server lets Claude read your Erlin data and run Erlin actions for you. You ask a question in Claude, Claude picks the right Erlin tool, and the answer comes back with real numbers from your account. MCP (Model Context Protocol) is the open standard Claude uses to connect to outside tools.

Erlin gives Claude 45 tools and 5 slash commands. They cover your tracked websites and brand profile, collections and prompts, visibility checks and their results, competitors, Google Analytics 4 and Google Search Console data, and content decay, keyword cannibalization, recommendations, Core Web Vitals and your branded search split. The MCP server does not store your data: it passes each request to your Erlin account and returns the result to Claude.

MCP comes with the Standard, Pro and Enterprise plans. Free and Starter do not include it. You also need Claude: the Claude Desktop app, Claude on the web, or Claude Code.

Start with a question about existing data

After connecting Erlin, ask Claude which websites you have in Erlin. You never type a tool name: Claude chooses the tool from what you asked.

Useful requests include:

"What is my AI visibility for the last 7 days?"
"How are we doing on ChatGPT compared to Perplexity?"
"Who are my top competitors by share of voice?"
"Which pages on acme.com are losing traffic?"

Slash commands are ready-made workflows. Type / in Claude to see them.

Command	What you get
`/erlin_weekly_visibility`	A 7-day visibility check-in with your top 10 competitors
`/erlin_competitor_deep_dive`	Your top 20 competitors, with suggestions for which to track
`/erlin_content_opportunities`	The pages to fix first, from content decay, recommendations and Search Console data
`/erlin_search_performance`	A Search Console review: top pages, branded split and index coverage
`/erlin_fanout_audit`	Which competitors win each variation of a topic in AI answers

Distinguish reading from changing

Every tool is either Read (it returns data) or Write (it changes something or starts work). Listing websites or reading a stored answer is a read. Creating or deleting prompts and collections, starting visibility checks and syncing analytics are writes. Review what Claude proposes before you allow a change.

Starting work is not the same as a finished result. Visibility checks and syncs run in the background, and Claude checks their status before reading the results.

Two paths show reading a previously collected answer or starting a new check and reviewing its status. — Read a collected answer, or start a check to collect a new one.

Understand the available scope

Visibility checks through the MCP server run on three AI platforms: ChatGPT, Gemini and Perplexity. Each check records whether your brand was mentioned, where it was ranked, and which sources the AI cited. Plans with more platforms track the others, such as Claude and Google AI Mode, in the Erlin app only.

Analytics answers need Google Analytics or Search Console connected in Erlin. When data is missing, Erlin says so rather than filling in an estimate.

Continue with visibility tools, prompt tools, analytics tools and credits and limits.