# Mentionary > Mentionary measures how often AI assistants recommend a brand when buyers ask unbranded questions. It reads a website's homepage, writes the questions buyers ask AI (for example "best CRM for small agencies"), asks ChatGPT, Gemini and Claude each one, and reports the mention rate per engine, the competitors named instead, the sources cited and a prioritised list of fixes. Engines measured on every plan: ChatGPT, Gemini and Claude. Google AI Overviews is built and running in pilot on selected workspaces. Perplexity is next. ## MCP connector - Connector URL (remote MCP, Streamable HTTP): https://platypus.mentionary.ai/mcp - Authentication: OAuth 2.1 with dynamic client registration and PKCE. The user signs in to their Mentionary account and approves the connection. There is no API key. - OAuth discovery: https://platypus.mentionary.ai/.well-known/oauth-authorization-server - Protected resource metadata: https://platypus.mentionary.ai/.well-known/oauth-protected-resource/mcp - Add to Claude: https://claude.ai/customize/connectors?modal=add-custom-connector&connectorName=Mentionary&connectorUrl=https%3A%2F%2Fplatypus.mentionary.ai%2Fmcp - ChatGPT (web): open https://chatgpt.com/plugins, select the plus button, choose "Add custom MCP server", enter the connector URL as the Server URL, choose OAuth, then "Create as a plugin". In a chat, type @ and select Mentionary. - Claude Code: `claude mcp add --transport http mentionary https://platypus.mentionary.ai/mcp` - Web app sign-up: https://app.mentionary.ai/auth/sign-up ## Tool flow A workspace tracks one website that the user owns or manages. Competitors are added to that workspace. A typical first run calls the tools in this order: 1. `create_workspace` (write): create a workspace for the user's domain and start the homepage analysis. 2. `get_setup_status` (read): poll until the homepage analysis has finished. Returns `next_step`, a plain sentence on what to call next, and analysis progress (answers checked of total). 3. `get_business_profile` (read): the detected value proposition, buyer personas and suggested competitors. 4. `confirm_business_profile` (write): save the value proposition and the personas to track. 5. Buyer questions, either way: - Written by you: `get_question_guidelines` (read) returns the business profile, personas, competitors, existing questions, free slots, question types by funnel stage, a default mix and the rules. By default write mostly top-of-funnel questions, plus questions about alternatives to competitors where that is how the brand wins customers; keep the brand's taglines and coined terms out of unbranded questions. Show the list to the person you work for, then save it with `add_queries` (write). Only questions that are too short, too long or already tracked are rejected. A question naming the brand is tracked as branded; one naming a competitor is tracked as a competitor question with its own rate, and that competitor is not counted against the brand on it. The response has a `summary`, `needs_attention` and, per question, its status, how it is counted and a `suggested_action` (null when nothing needs changing). - Written by Mentionary: `generate_queries` (write) writes the buyer questions and activates the recommended set up to the plan's cap (mostly unbranded, spread across personas, at most a fifth naming the brand). Can take a couple of minutes. 6. `list_queries` (read): list the questions and whether each is active. 7. `set_queries_active` (write): turn tracking on or off per question, within the plan's cap. `activate_recommended_queries` (write) fills free slots with the recommended questions; use it when coverage is `thin`. 8. `start_ai_visibility_analysis` (write): ask ChatGPT, Gemini and Claude every active question. Runs in the background, about a minute per five active questions (about 2 minutes on the Free plan). Returns `estimated_seconds` and `check_back_in_seconds`; partial results can be read while it runs. 9. `get_visibility_brief` (read): the story, worked out: a headline on unbranded questions, a confidence level, key findings with numbers, the biggest risk and opportunity, and chart-ready data. Start summaries, reports and decks here. `get_ai_visibility_results` (read): based on the latest answer from each engine to each active question. Mention rate overall and per engine (ChatGPT, Gemini, Claude; pilot engines reported separately), competitors named instead, most cited sources, sources cited where a competitor is named and the brand is not, the change on the same questions since the previous answers, and a per-question breakdown with persona and cited sources. `is_partial` is true while an analysis is still running. 10. `set_competitors` (write, destructive): replace the competitor list and target markets. Past mention history of removed competitors is kept. 11. `get_citation_sources` (read): the sources engines cite, as in the web app's citations dashboard: totals for the site's own pages and third-party sites and per engine, the site's own pages that get cited, the third-party domains cited most with authority score (0-100) and top URLs, and a weekly trend. Last 1-12 weeks, sortable by citations or authority. 12. `get_recommended_actions` (read): measured fixes with their evidence (sites to earn a mention on, questions to answer), site recommendations from the homepage analysis, and content gaps. 13. `list_workspaces` (read): list the workspaces the account can access. Call this first for a returning user. 14. `complete_setup` (write): finish workspace setup. ## MCP prompts The server also offers ready-made prompts (slash commands in Claude), each taking the domain: `ai_visibility_check`, `landing_page_for_gap`, `weekly_check`, `sources_to_win`, `citation_report`, `board_report`, `fixes_to_tickets`, `client_brief`. ## Notes for agents - The person you work for signs in once through OAuth. A new account starts on the Free plan, no card needed. - Tools identify a workspace by its domain (`workspace: "acme.com"`). Call `list_workspaces` first; `create_workspace` returns the existing workspace if the account already tracks the domain. - Question ids are opaque codes such as `q_3HYPJKZ`; pass them back as returned. - Analyses run in the background. Poll `get_setup_status` until the stage you need has finished, then read results. - Results cover ChatGPT, Gemini and Claude. Do not report other engines as measured. - `get_setup_status`, `get_visibility_brief` and `get_ai_visibility_results` include a `checkup`: ranked suggestions to improve the measurement (too few questions, one persona, unmeasured questions, stale results, no confirmed competitors, too many branded questions, lookalike brands cited), each naming the tool that fixes it. Offer to act on the high-priority ones. - Personas are free text. Add buyers with `add_personas` (name, optional role and pain points), or write questions for them in `add_queries`, which creates unknown personas. - Any member of a workspace can start an analysis. If every active question already has results, nothing runs unless `rerun` is true. - Lead with the unbranded mention rate (questions that do not name the brand). Branded questions inflate the overall rate. Competitor questions (such as 'free alternatives to X') have their own rate. Check `confidence` and `question_coverage` before drawing conclusions. ## Playbooks Chained prompts for common jobs, run in order in one conversation. Full prompts are in https://mentionary.ai/llms-full.txt. - Building a page: map which competitors the engines name for your category and the sources they cite, then draft a landing page for the question you trail on most. - Improving a live site: see what moved since the last run, rank the third-party sites cited where competitors are named and you are not, then turn the fixes into tickets. - Reporting to the board: a one-page AI visibility report (mention rate per engine, competitors named instead, three fixes), then a five-slide outline. - Running it for clients: a brief per client site before the call, then a status line across every client workspace. ## Pricing - Free: $0, no card. 1 website, 10 tracked buyer questions, 1 full analysis across the three engines, 2 competitors. - Starter: $29 a month. 25 questions, 150 answer checks a month, 2 competitors. - Pro: $99 a month. 100 questions, 1,000 answer checks a month, 10 competitors, automatic weekly re-analysis. - Agency: $299 a month. 500 questions, 5,000 answer checks a month, unlimited competitors, automatic weekly re-analysis. Paid plans have a 14-day trial that starts at checkout. ## Key pages - [Home](https://mentionary.ai/): what Mentionary measures and why - [MCP connector](https://mentionary.ai/mcp): connection steps for Claude, ChatGPT, Claude Code and other MCP clients - [Claude setup](https://mentionary.ai/mcp/claude), [ChatGPT setup](https://mentionary.ai/mcp/chatgpt), [Claude Code setup](https://mentionary.ai/mcp/claude-code) - [Use cases](https://mentionary.ai/use-cases): six prompts and what each returns - [Methodology](https://mentionary.ai/methodology): how the measurement works and its limits - [Glossary](https://mentionary.ai/glossary): definitions of mention rate, buyer question, answer check and related terms - [Why asking ChatGPT about your brand is not a measurement](https://mentionary.ai/guides/why-asking-chatgpt-about-your-brand-is-not-a-measurement) - [Pricing](https://mentionary.ai/pricing): plans and limits - [RE24.energy case study](https://mentionary.ai/customers/re24-energy): +72% AI citations in 87 days - [FAQs](https://mentionary.ai/faqs) - [Full guide for agents](https://mentionary.ai/llms-full.txt): every prompt, tool, plan and FAQ in one file - [Research](https://mentionary.ai/research)