Connect Refora to Claude and ChatGPT

Refora's MCP server lets AI assistants read your Refora data (AI visibility, share of voice, competitors, citations, sentiment, prompt performance and the AI answers themselves) so you can ask questions about it in plain language. The connection is read-only: an assistant can never change anything in your Refora account.

Server URL: https://mcp.refora.ai/mcp

Watch the demo

Before you start

  • You need a Refora account on a paid plan or an active trial.
  • You can connect any Refora organization you are a member of. You choose the organization when you sign in, and each connection reads only that organization.
  • Signing in uses your normal Refora email and password.

Claude (claude.ai, Claude Desktop and the Claude mobile apps)

  1. Open Customize → Connectors (Settings → Connectors links there) and choose Add → Add custom connector.
  2. Name: Refora. MCP server URL: https://mcp.refora.ai/mcp. Choose Continue. Claude checks the server and detects the sign-in; keep the detected options and choose Add.
  3. Choose Connect. Sign in to Refora, pick the organization, and choose Allow.
  4. In a chat, open the tools menu and make sure Refora is turned on. Then ask a question (see the examples below).

On Claude Team and Enterprise, an Owner adds the connector once under Organization settings → Connectors. Each member then chooses Connect and signs in with their own Refora account.

ChatGPT

  1. In ChatGPT on the web, open Plugins in the sidebar and choose Add → Add custom MCP server.
  2. Name: Refora. Server URL: https://mcp.refora.ai/mcp. Keep Authentication on OAuth, tick I understand and want to continue, and choose Create as a plugin.
  3. Choose Connect Refora. Sign in to Refora, pick the organization, and choose Allow.
  4. Open Refora under Plugins and choose Try in chat. Then ask a question about your Refora data.

On ChatGPT Business and Enterprise, an admin may need to allow the app for your workspace first.

Claude Code and other MCP clients

Add Refora as a remote (Streamable HTTP) MCP server. In Claude Code:

claude mcp add --transport http refora https://mcp.refora.ai/mcp
claude mcp login refora

Cursor, VS Code and other MCP-compatible tools work the same way: add an HTTP MCP server with the URL above, and sign in when asked.

What you can ask

  • "How visible is my brand in AI answers this month, and which AI model mentions us least?"
  • "Which competitors does AI mention more than us, and on which questions?"
  • "Which websites do ChatGPT and Perplexity cite when they answer questions about our category?"
  • "How has our AI visibility changed over the last 90 days, week by week?"
  • "How do AI assistants describe our brand: positive, neutral or negative?"
  • "What did ChatGPT actually say when asked about our category, and which sources did it cite?"
  • "Show me last month's AI answers that don't mention us."

The tools

All tools are read-only.

ToolWhat it returns
list_workspacesThe workspaces in the organization you connected. Start here.
get_visibility_scoreThe share of AI answers that mention your brand, overall and per AI model, with the change from the previous period.
get_visibility_trendYour visibility over time, by day, week or month.
get_share_of_voiceYour share of all brand mentions in AI answers, and which brands hold the rest.
get_competitor_comparisonYour tracked competitors ranked by visibility.
get_competitor_visibility_gapThe questions where a competitor is mentioned more often than you, and the sources AI cites in those answers.
get_citation_sourcesThe domains and pages AI models cite in answers to your prompts.
get_top_cited_urls_for_topicThe most-cited pages for one topic.
get_top_cited_domainsThe most-cited domains across all workspaces in the organization.
get_prompt_performanceOne prompt's visibility, per-model results and cited sources.
get_ai_answersThe answers themselves: what each AI model said to your prompts, with the brands each answer mentions and the sources it cites. Filter by prompt, AI model, brand or period.
get_sentiment_breakdownHow AI models describe a brand, with short example excerpts.

Reading the numbers

  • Visibility is the share of AI answers that mention your brand. Runs that returned no answer are not counted.
  • When there is nothing to measure in a period (no answers yet, or no brand set up in the workspace), the result is empty rather than 0%.
  • Results cover the last 30 days by default, and changes compare with the 30 days before. You can ask for any period up to 90 days.
  • Refora asks its prompts to the AI models on a schedule, so periods shorter than a week can be empty.
  • AI answers come ten at a time, newest first, and long answers are shortened. Ask the assistant for the next ones, or to show one answer in full.

Disconnecting

  • Remove the connector in Claude or ChatGPT at any time.
  • If someone is removed from your Refora organization, their connections stop working immediately.
  • To have a connection revoked on our side, email support@refora.ai.

Limits

Each connection can make up to 100 requests per hour.

Troubleshooting

  • "None of your organizations can connect AI apps yet": connecting AI assistants is included in paid plans and trials. Upgrade, or contact us.
  • The sign-in page says it has expired: sign-in links last ten minutes. Go back to Claude or ChatGPT and start the connection again.
  • The assistant stops working after a while: sign-ins renew automatically. If it still fails, remove the connector and add it again.

Privacy and security

When you ask an assistant a question, Refora sends it only the data needed to answer, from the organization you connected. That data is then processed by the assistant's provider (for example Anthropic or OpenAI) under its own terms. Read more in our Privacy Policy.

Questions: support@refora.ai · Security issues: security@refora.ai