NameIO
See which shopping questions AI assistants answer with a competitor instead of your store, and draft page fixes.
Размещённый MCP-сервер
npx add-mcp 'https://app.getnameio.com/mcp'Устанавливается в Claude Code, Codex, Cursor и другие
Документация
Connect the AI assistant you already use to NameIO, then ask it to do your research from there: keyword research, Google search results, domain and backlink analysis, rank tracking, AI visibility checks, and your Google Search Console and Google Analytics data. It reads your NameIO projects and, when you allow changes, can save keywords, questions to track, reports, project setup and draft page fixes back to them, where they show up in the app.
NameIO is a remote MCP server at https://app.getnameio.com/mcp. You add the server to your client, sign in, and choose whether the assistant may make changes.
NameIO doesn't run its own AI model to reason or write for you. Your assistant does that; NameIO supplies the data, including the AI answers it collects for your checks, and stores what the assistant saves.
Paste the setup prompt
NameIO's setup prompt tells your agent how to add the server, sign you in and check the connection.
- In NameIO, open AI assistant from the Your AI assistant group in the sidebar.
- Click Copy setup prompt.
- Paste it into your agent and follow its instructions.
- Approve the connection when NameIO asks. See "Approve the connection" below.
The agent finishes by running whoami and list_projects to check the connection. The setup prompt asks it not to create projects or run paid research during setup.
The same prompt is on the last onboarding step, Connect your AI assistant (optional: See my results goes straight to the dashboard), which shows Connected once an assistant connects and then continues to the dashboard. That step, under How to connect your assistant, also shows the server URL with a copy button and the Claude steps. On the AI assistant page, the server URL, the Claude steps and the Gemini app steps are under Steps for Claude, Gemini and Copilot Studio (closed until you open it). Inside it, the Gemini Enterprise and Microsoft 365 Copilot steps are under For your IT team. Until an assistant connects, the dashboard's Write fixes with ChatGPT or Claude card links to the AI assistant page.
In Claude (claude.ai and Claude Desktop), ChatGPT and the Gemini app, you add the connector yourself, using the steps for your client below rather than the setup prompt's. In Microsoft 365 Copilot it's added in Copilot Studio; see that section.
Claude Code
- Add the server:
This uses Claude Code's default local scope. Addclaude mcp add --transport http nameio https://app.getnameio.com/mcp--scope userto use it across all your projects, or--scope projectto write it to the project's.mcp.json. - In Claude Code, run
/mcp, select nameio and follow the browser steps to sign in.claude mcp login nameiodoes the same. - Approve the connection in NameIO.
To use an API key instead of signing in, add it as a header:
claude mcp add --transport http nameio https://app.getnameio.com/mcp --header "Authorization: Bearer nio_your_key"
Don't combine a key with --scope project, which would write it into a file in your repository.
Codex CLI
- Add the server:
codex mcp add nameio --url https://app.getnameio.com/mcp - Start sign-in, then sign in to NameIO in your browser and approve the connection:
codex mcp login nameio - In the Codex TUI, run
/mcpand check that nameio is listed.
Codex saves the server as a [mcp_servers.nameio] table in its configuration, which the Codex IDE extension also uses. To use an API key instead of signing in, keep the key in an environment variable and name that variable in bearer_token_env_var:
[mcp_servers.nameio]
url = "https://app.getnameio.com/mcp"
bearer_token_env_var = "NAMEIO_API_KEY"
Cursor
Add NameIO to Cursor opens Cursor with the server filled in; confirm it, then sign in to NameIO when asked. Or add it by hand:
- Open
~/.cursor/mcp.jsonto add NameIO to all your projects, or.cursor/mcp.jsonin a project to add it there only. - Add a
nameioentry next to any servers already listed:{ "mcpServers": { "nameio": { "url": "https://app.getnameio.com/mcp" } } } - Cursor supports OAuth for servers that require it. Sign in to NameIO when asked and approve the connection.
You manage MCP servers from Cursor's Customize page. To use an API key instead of signing in, read it from an environment variable in a header:
{
"mcpServers": {
"nameio": {
"url": "https://app.getnameio.com/mcp",
"headers": {
"Authorization": "Bearer ${env:NAMEIO_API_KEY}"
}
}
}
}
Claude (claude.ai and Claude Desktop)
Custom connectors work in Claude, Cowork and Claude Desktop on the Free, Pro, Max, Team and Enterprise plans. On Free you can add one custom connector. Claude reaches NameIO from Anthropic's cloud, not from your computer.
On an individual plan:
- Open Customize > Connectors, click + Add, then Add custom connector.
- Enter the name NameIO and the URL
https://app.getnameio.com/mcp, click Continue, review the detected authentication settings and click Continue again. - For sign-in, choose Sign in now or Sign in when needed.
- For the OAuth client, keep Use Claude's published identity (Recommended). NameIO's consent page then shows which site confirmed the app's identity. Register automatically also works.
- Click Add.
- When Claude sends you to NameIO (during setup with Sign in now, or the first time you use the connector with Sign in when needed), sign in and approve the connection.
- In a chat, click + in the lower left, open Connectors and turn on NameIO.
On an individual plan, if you can't use sign-in and your dialog shows Request headers (a beta that only some accounts have), choose No sign in, add an authorization header and enter Bearer followed by your NameIO API key.
On Team and Enterprise, an Owner adds the connector under Organization settings > Connectors (Add, then Custom, then Web) with the URL above; either OAuth client choice works. Each member then clicks Connect and signs in to their own NameIO account.
ChatGPT
OpenAI offers full custom MCP servers on the Business and Enterprise/Edu plans; Pro accounts can connect one with read and fetch permissions only. On Business, only admins and owners can add them. On Enterprise/Edu, admins and owners can, and so can members given the Create plugins with MCPs permission. You add custom MCP servers in ChatGPT on the web.
- On chatgpt.com, open Plugins (chatgpt.com/plugins).
- Click the plus button, then Add custom MCP server.
- Enter the name NameIO. The description is optional.
- Under Connection, enter
https://app.getnameio.com/mcp. - For authentication, choose OAuth.
- Tick I understand and want to continue, then click Create as a plugin.
- Review the tools ChatGPT found and install the plugin. When ChatGPT sends you to NameIO, sign in and approve the connection.
- Start a new conversation, type @, select NameIO and send: "Use NameIO to check my connection and list my projects."
By default, ChatGPT asks you to confirm actions that can make changes, and it also asks for many NameIO tools that only read data or run lookups. You can approve a tool for the rest of a conversation; a new conversation asks again.
Gemini
The Gemini app (personal Google Account). You add NameIO as a custom app in the Gemini web app; it then works in Gemini chats on the web and in the Gemini mobile app. Google's conditions: a personal Google Account, 18 or over, in the US, in English, with Keep Activity on. Work and school accounts can't add custom apps; use the Gemini Enterprise steps below. If there is no + Custom button on the Connected Apps page, your account can't add custom apps yet.
- On gemini.google.com, open Customize → Connectors (gemini.google.com/apps) and click + Custom. Google's help page calls the same page Settings → Connected Apps.
- In Connect to an MCP server, enter
https://app.getnameio.com/mcpas the MCP Server URL, leave Additional settings alone and click Next. - Sign in to NameIO and approve the connection.
- In a chat, type @, select NameIO and send: "Use NameIO to check my connection and list my projects."
Gemini asks you to confirm each write action. To disconnect, open More details on the app under Custom apps and choose Disconnect or Remove app; that also unlinks NameIO from your Google Account. Google's help: Connect & manage custom apps.
Gemini CLI and Antigravity CLI. Both take a remote Streamable HTTP server with OAuth. Add "nameio": { "httpUrl": "https://app.getnameio.com/mcp" } under mcpServers in ~/.gemini/settings.json (Antigravity CLI imports that file and has its own MCP settings), or use the CLI's mcp add command for a remote HTTP server, then run /mcp auth nameio and sign in when the browser opens. Since June 2026 Gemini CLI serves only Google Cloud and API-key accounts; individual accounts use Antigravity CLI.
Gemini Enterprise (work account). A team administrator adds NameIO as an MCP server; members then sign in to NameIO once and use it in chat and in agents. Google's form asks for a client ID and secret, which you create on NameIO's AI assistant page.
- Business edition: at business.gemini.google, open Settings & help, pick the team, then Manage team → Connected apps → Add MCP Server. Cloud edition: in the Google Cloud console, Gemini Enterprise → Data stores → Create data store → Custom MCP Server.
- Server URL
https://app.getnameio.com/mcp, name NameIO. Authentication OAuth 2.0: Authorization URLhttps://app.getnameio.com/api/auth/oauth2/authorize, Token URLhttps://app.getnameio.com/api/auth/oauth2/token, Scopesoffline_access mcp mcp:write(oroffline_access mcpfor read-only access). Cloud edition: Enable PKCE Support can stay off; NameIO issues a client secret. - Client ID and Client Secret: on NameIO's AI assistant page, under Using Gemini Enterprise at work, click Create client ID and secret and paste them in. They are shown once and work only with Gemini Enterprise. Copy the brief for IT gives the same steps as one paragraph, without the secret.
- Click Verify Auth, then Continue, pick the location, name the connector and click Add. The connector starts disabled: enable it for the team.
- Members sign in to NameIO when Gemini asks (each sees only their own NameIO projects). In an agent, open Tools & Integrations and add NameIO.
Google's help: Set up your custom MCP server connection (Business edition) and Set up your custom MCP server data store (Cloud edition). Both call the feature pre-GA.
Microsoft 365 Copilot
The Microsoft 365 Copilot chat can't add an MCP server itself. NameIO is added to an agent in Copilot Studio, by you if you have a Copilot Studio license or by whoever builds agents for your company, and that agent is then published to Microsoft 365 Copilot. Copilot Studio connects to remote Streamable HTTP servers like NameIO and supports both of NameIO's sign-in methods.
- Open the agent in Copilot Studio (or create one), then Tools → Add a tool → New tool → Model Context Protocol.
- Enter the server name NameIO, a short description the agent uses to decide when to call it, for example "Shows which buying questions AI assistants answer with the store's competitors instead of the store, and drafts page fixes", and the server URL
https://app.getnameio.com/mcp. - For authentication, choose OAuth 2.0 and Dynamic discovery, then sign in to NameIO when asked. If Dynamic discovery fails, choose API key, type Header, name
x-api-key, and paste a key from Settings → API keys when the connection asks for it. A key has full read-and-change access; with OAuth each person can choose read-only when they approve the connection. - Click Create, then Next, create the connection (sign in to NameIO and approve it) and Add to agent. Keep the tool on end-user credentials, so each person signs in to their own NameIO account rather than sharing yours, and keep generative orchestration on.
- Test the agent with "Use NameIO to check my connection and list my projects", then publish it. Publishing to everyone in the company usually needs an administrator's approval, and a Power Platform data policy can block custom tools until an administrator allows this one.
If someone else manages Copilot for your company, the setup prompt and the AI assistant page include a short brief you can forward to them.
Microsoft's own steps: Connect your agent to an existing MCP server.
Other MCP clients
Add NameIO as a remote Streamable HTTP MCP server at https://app.getnameio.com/mcp using your client's current documentation, and sign in with OAuth. If the client can't complete the sign-in, use an API key.
Approve the connection
When your client starts sign-in, NameIO opens in your browser. If you aren't signed in, click Continue with Google first. NameIO then shows the consent page, Authorize MCP access.
- Check the details. An app with a published identity shows the site that confirmed it (for Claude Code, claude.ai) and the name it calls itself; otherwise the app's name comes from the app and isn't verified. Then check where access is sent and the account you're signed in as. Only continue if you started this connection yourself.
- Read any warning. If access goes to a program on your computer, any program running there could receive it. If it goes to an app on your device through a custom link type, any app registered for that link type could receive it.
- Review the access. Read your NameIO data and run lookups is always included: projects, AI visibility results, keywords, reports and audits. Lookups use your organization's credits.
- If the app asked to make changes, Make changes is listed with a switch beside it, on by default. Turn it off to connect read-only.
- Click Authorize.
Check the connection
Ask your assistant: "Use NameIO to check my connection and list my projects." It runs whoami and list_projects, which use no credits.
whoami shows your account email, the access you granted (a read-only connection doesn't include mcp:write) and the credits left in your current organization.
In NameIO, Settings → Personal → Connected AI assistants lists each assistant you connected through sign-in, with its access (Read and change or Read only) and the date it connected. An app with a published identity is listed by the site it was verified against. For API keys, the API keys table shows when each key was last used.
If an assistant you connected through sign-in goes unused for 30 days, sign in again from your client.
Read only or read and change
A read-only connection can read your saved data and run lookups, which still use credits. It can't create, change or delete anything saved, or start checks and audits.
With the Make changes switch on, the assistant can also:
- create projects and save a project's website setup and context
- save and remove saved keywords
- save AI visibility questions, set the AI check schedule and start a one-off AI check
- create rank trackers, add, remove or pin tracked keywords, and run rank checks
- start and delete site audits
- save, share, unshare and delete reports
- save and delete report templates
- save a draft fix for one of your store's pages. A fix is only a draft: you review it, then paste it into your store yourself. Your assistant can't change your store. See "Draft a fix for my store" below.
- record its own read of how a stored AI answer describes each brand it names (
record_answer_tone). See "Read the tone of answers" below. - set the funnel stage of a question (
set_question_stage). See "Set a question's funnel stage" below. - give a question other wordings (
set_question_phrasings). See "Give a question other wordings" below.
To turn a read-only connection into one that can make changes, disconnect it under Settings → Personal → Connected AI assistants, then connect again from your client and leave the Make changes switch on.
If an app asks only for read-only access, the consent page says so and shows no switch.
Use an API key
Use an API key when your client can't complete the OAuth sign-in.
- In NameIO, open Settings from the account menu at the bottom of the sidebar (https://app.getnameio.com/settings).
- On the Personal tab, under API keys, click Create API key.
- Enter a name of up to 32 characters, such as "Claude Code on laptop", and click Create.
- Copy the key. It starts with
nio_and isn't shown again. - Put the key in your client's secret settings or an environment variable, never in a chat or a repository. Have the client send it to
https://app.getnameio.com/mcpasAuthorization: Bearer nio_.... Anx-api-key: nio_...header also works.
A key has full read-and-change access. It works only with the MCP server and doesn't sign you in to the NameIO app. It doesn't expire; it works until you revoke it. The API keys table lists each key's name, its first characters, when it was created and when it was last used.
Organizations
One connection covers every NameIO organization you belong to, whether you connect through sign-in or an API key. When you have more than one, list_projects labels each project with its organization. If you leave an organization, the connection loses access to its projects; if you leave every organization, it stops working.
What to ask first
If you have no projects yet, your assistant can ask for your website and create a project. Run your store setup in the app (the dashboard's Set up your store button opens the form on Prompt Tracking); the assistant can't run it. After that, an assistant with Make changes on can add, pause or edit questions. A good first task for it is to fill in the project's context: it studies your website itself and saves a business overview and up to 5 competitors. Creating and saving need a connection with Make changes on.
You can list the sub-brands and flagship products your store and your competitors go by, in the store setup form or on the project's Context page, or ask your assistant to save them (it needs Make changes on). An AI answer that uses one of those names counts as naming that brand. A common word or a number on its own, or a name that's in your questions, can't be used.
Then try: "Use NameIO to show which buying questions AI assistants answer with my competitors instead of me."
To see how AI has described your store over time, try: "Use NameIO to show how AI described my store against my competitors over the last 12 months." Your assistant reads the counts month by month for the last 6, 12 or 24 months (the tool get_ai_reputation_history, on NameIO at app.getnameio.com), as plain counts, never a score, and gives a change from the month before or the year before only when both months have enough answers to compare. See How AI describes you.
For several sites, ask your assistant to set up a separate project for each website you list, and to list your existing projects first and reuse matches, so it doesn't create duplicates.
AI visibility checks cover ChatGPT, Gemini, Google AI Overviews and Perplexity. Grok, Meta AI and Kimi are planned for Pro and not offered yet. For a ChatGPT answer the assistant can also read the product cards it showed, the pages ChatGPT read while searching, and the searches it ran. For a Google AI Overviews answer it can read the products linked inside the overview and the results page's popular products and sponsored listings, each saying where it came from. For each competitor, the results carry how the answers describe your brand next to it (better, even or worse, over the answers naming both), with each brand's caution words by kind and what answers to questions with your store's name say about it, and each answer carries the words it counted. The get_agent_readiness tool reads whether each site you track publishes a checkout profile that AI assistants can buy through. The get_site_readiness tool reads your Site readiness page; see Check your site for AI search. The answers are a sample: they vary from check to check and can differ from what you see in each app. The quick setup in NameIO onboarding tracks ChatGPT, Gemini and Google AI Overviews on the free plan, and those and Perplexity on the Pro plan. See AI visibility.
Read a number over time
To see how one number moved, try: "Use NameIO to show how many AI answers named my store in each check over the last 28 days." Your assistant reads one count at a time (the tool get_metric_series): over the last 7, 28 or 90 days, one point for each full check (a quick Run now of a few questions isn't one), or over the last 6, 12 or 24 months, one point for each month. Each point is a count such as "7 of 12 answers", with the name and definition from the Metrics table. The counts it reads:
- Your store in the answers:
vis.answered(Answered),vis.named(Named),vis.cited(Linked) andvis.namedFirst(Named first). - You against a competitor:
rep.compared(Named with a competitor),rep.better,rep.evenandrep.worse(you described better, the same, the competitor described better), andrep.betterShare(the change in answers that describe you better, over months only). With a competitor's website, it reads you against that competitor; without one, against every competitor added up. - Words about your store:
rep.withCaution(caution words, every kind together),rep.serious(serious words) andrep.role(given a role). - Product cards:
cards.shownandcards.rating. - Site checks:
site.passedandsite.failed, one point for each site check of your store, or of a competitor's site with its website, with the days NameIO changed what it checks.
With a competitor's website, the counts about your store are that competitor's own instead. The counts are for questions without your store's name. For caution words and serious words, your assistant can also read the questions with your store's name (the scope branded). For one question, try: "Use NameIO to show how often AI answers to my question about trail shoes named my store over the last 90 days." Your assistant reads that question's answers alone (the scope question: Named, Linked or Named first), one point for each check that asked it, as the question's page shows them.
Your assistant gets a change from the earlier period only when both periods have at least 10 answers, asked the same questions and moved by at least 2 answers or 2 points; otherwise it gets the reason. A change in a count of answers or cards is per check ("+3 answers per check"), so a period with more checks doesn't look bigger. It works on NameIO at app.getnameio.com and on a NameIO you run yourself, uses no credits and starts no check.
Other tools give a few more numbers:
get_ai_visibility_resultsalso gives the competitor your dashboard compares you with first and why, that competitor on each engine, the check before, and the serious words for each kind of question.get_ai_visibility_trendalso gives each rate as a count ("4 of 10" questions and engines) and says when there are too few to compare; its percentage for how answers describe you against competitors is kept for now, andget_metric_seriesgives the same comparison as counts.get_ai_reputation_historyalso names the count behind each change and says which months have too few comparisons.get_fix_impactalso names the count it reports: answers that name your store or link to your site. It starts with the count the fix should move (target), as below.
Read your funnel
To see where AI loses your shoppers, try: "Use NameIO to read my AI funnel." Your assistant reads your funnel (the tool get_funnel). For each funnel stage (Awareness, Consideration and Conversion) it gets three numbers: how often AI names your store, your share of brand mentions (your part of all the brands AI names), and your place among those brands. Each number comes with a likely range, such as "likely 22% to 41%", or with a count when there are too few answers. A place is a place, such as "2nd to 4th of 6 brands named", never a percentage. It is counted among your store and at most 9 other brands, the ones AI ranked often enough, and for now it is a rough guide. Your assistant also gets the changes against the 4 and the 13 weeks before, in points with a likely range, your lowest step (or why no step is named yet), the questions that moved most and the brand list behind the shares, with how many names count, how many are not yet sorted and how often answers name a retailer. Ask it to read your funnel before it proposes a fix. With a stage, get_metric_series gives the same numbers week by week (vis.named, vis.mentionShare and rep.rank): over every question asked, or only over the questions asked through most of those weeks (the basket same), so a new question doesn't move the line. It works on NameIO at app.getnameio.com and on a NameIO you run yourself, uses no credits and starts no check.
To go from your funnel to what to do next, try: "Use NameIO to read my AI funnel and tell me which step to work on first." Your assistant works in this order:
- It reads your funnel and gives your lowest step as NameIO words it, or why no step is named yet. It gives each number as NameIO gives it: a count where there are too few answers for a percentage, and a change in points with its likely range.
- Where a step has too few answers, it suggests 3 to 5 new questions for that step, without your store's name, with what they add to the cost of a check. On the Pro plan it can also suggest two other wordings for each of that step's questions. It saves them only when you say yes, and saving needs Make changes on.
- Where a change can't be told yet, more weeks narrow it, and new questions don't.
- It opens the fix brief for a row of What to do next whose questions are in that step, and tells you the smallest change that fix could show (see Draft a fix for my store; on NameIO at app.getnameio.com). Nothing is published.
How NameIO counts each number, what one answer is and how a change is called are on the Methodology page.
Set a question's funnel stage
Each question you track counts in one funnel stage: Awareness, Consideration or Conversion. NameIO picks the stage from the question's words, and each starter question has a fixed stage. To move one, try: "Use NameIO to move my question about returns to Conversion." Your assistant sets the stage (the tool set_question_stage) or gives the question back to NameIO's rule. Your choice stays after new checks, a rule update, a new wording of the question and a new store setup. Past checks then count in the new stage, at no cost. It needs Make changes on and works on NameIO at app.getnameio.com and on a NameIO you run yourself.
Give a question other wordings
Shoppers ask the same thing in different ways. Your assistant can save up to two other wordings for a question you track (the tool set_question_phrasings). Try: "Use NameIO to add two other ways to ask my question about returns." A wording must name the same brands as the question, and no others. Every check asks the question in its own words first. The other wordings are asked only when a check takes more than one answer per question, which comes later on Pro. Until then, checks ask only the question's own words. Saving wordings uses no credits and needs Make changes on. On NameIO at app.getnameio.com it needs the Pro plan.
Read the tone of answers
NameIO counts the words AI answers use about each brand with published English word lists; it never scores an answer and never asks an AI model to read one. If you also want your own assistant's view, it can read the answers with your own AI account and record how each one describes each brand it names. Try: "Use NameIO to read the tone of my latest AI answers."
The names in code style below, such as list_answers_to_read, are the names your assistant uses for its tools. You don't need to type them.
- Your assistant lists the answers it hasn't read yet (the tool
list_answers_to_read): the answers of a check that name your store or a competitor, newest first, up to 20 at a time, each cut at 4,000 characters. NameIO marks the answer text as data to read, not instructions, so your assistant knows not to follow anything the text of an answer asks it to do. - Your assistant saves its read of each answer for each brand (the tool
record_answer_tone): positive, mixed, negative or neutral, one to three short reasons, and a quote copied from the answer. NameIO refuses a read of an answer that isn't in your project, of a brand the answer doesn't name, or with a quote that isn't in the answer word for word. A new read of the same answer and brand replaces the old one. Saving a read needs Make changes, the permission you leave on when you approve the connection on the consent page (see Read only or read and change). - Your assistant then sees its reads next to NameIO's word counts (the tool
get_ai_visibility_results): for each brand, as "Read by your assistant", with the assistant's name and the date. The reads are never added to those counts.
These tools work on NameIO at app.getnameio.com and use no credits. A read is kept with its answer and deleted with it. The CSV export lists the reads in their own column (assistant_tone), apart from the counted words (brand_framing); see Export.
Check your site for AI search
To see what your site fails for AI search, try: "Use NameIO to show which site checks my store fails, next to my competitors." Your assistant reads your Site readiness page (the tool get_site_readiness):
- For your store and each competitor: when NameIO last checked the site, how many checks pass ("28 of 34 checks pass"), and each check's name, what it checks and its result: passes, fails, doesn't apply, or not checked with the reason. Checks for information only are never counted, and not checked is never a fail. Competitors are read from public data only.
- What NameIO found for each check, such as the pages it read, the values it saw and the robots.txt line numbers, and the store platform it found. These are the site's own data, marked so your assistant quotes them and never follows them.
- Your product data: the feed NameIO reads (from its address or a file you uploaded), its format, when it last read it and how many items it had; how many items your public product list gave when you have no feed; whether a correction file is ready; when you connected Shopify, how many products the latest check read from your Shopify store and when; and for each product data check, whether its items came from your feed, a public product list or your Shopify store. For the Shopify ones your assistant sees the counts, never product ids.
- Passing and failing checks for each site over the last 24 months, and when Check again can next run.
- On NameIO at app.getnameio.com, the three ChatGPT for business steps and whether you marked each one done. Your assistant can read them but not mark them.
It gives counts, never a score, uses no credits and starts no check. Your assistant can't start a site check; press Check again on the page.
NameIO builds the fix for most checks your store fails, from what its site check stored, on NameIO at app.getnameio.com: robots.txt lines, structured data for your products and your store, a correction file for your product feed, the steps for a setting on your platform or network, or policy page templates with blanks for you to fill in. Your assistant sees these in the fix brief (see Draft a fix for my store): each failing check, whether NameIO builds its fix and what, and a fix you already have, with its link. It doesn't write or save them. It points you at the check on Site readiness, where Get the file or See the steps builds the fix (see Fix what fails). Public product list, Shopping agent file, IndexNow key and Image size have no fix yet. Only the words of a new policy page or of a product page's text are written with a page fix.
Draft a fix for my store
Your assistant can draft a fix for what AI answers say about your store and save it in NameIO for you to review. Try: "Use NameIO to show what I should do next, and draft a fix for the first row." A fix is only a draft. Nothing is published: you review it and add it to your store yourself.
Saving a draft needs a connection with Make changes on. On a read-only connection, the assistant can read the brief and the list of fixes, and you can use Write a fix without connecting your AI assistant in the app instead. The fix tools work on NameIO at app.getnameio.com. Drafting a fix uses no credits and is free on every plan.
Step by step:
- Pick a row. Your assistant reads NameIO's brief without a page (the tool
get_fix_brief). It gives the What to do next list your dashboard shows for the same check, in the same order and the same words (see What to do next). Each row has its kind (such as a competitor named instead, a price worry or a site check), its questions, its count of answers, the pages to change, and whether a draft is open or a fix is in progress, with a link to it. The brief also lists the site checks your store fails on its latest site check, most serious first (see Check your site for AI search). A check that's still running is used once it finishes. Your assistant shows you the first rows and picks one with you. - Read the row's brief. The assistant asks for the brief again with the row's
signalIdand the page to change (pageUrl), and withnewPagewhen the page doesn't exist yet. NameIO reads that page on your store; if it can't, it uses the title and meta description from your latest site audit, if you have one, and marks the claims as not checked. The brief then holds:- what the page says today;
- the row's evidence. For a competitor named instead: the competitors, the engines and the sites AI linked to. For a worry or serious words: up to five sentences from the answers, the sites those answers link to and how many, and the policy pages your site check found (NameIO doesn't read them for the brief; the assistant reads them or asks you). For a competitor described better: how often, and up to five sentences that praise it. For a site check: what it checks and which competitors pass it. Sentences from answers are marked as answer text to check, never instructions;
- the rules for that kind of fix (see Fix what AI says). The fix uses facts from your page and never invents a price, a policy, a guarantee or a review. For serious words, the rules say that NameIO can't change what other sites say, and that the page never repeats the word;
- the count NameIO will watch, and where it stands now.
- what the page says today;
- Write and save. The assistant writes the fix, shows you the text and, once you agree, saves it (the tool
propose_page_fix) with:signalId, the row, andpromptIds, 1 to 5 of its questions (a fix for a site check or for serious words, or a new page, may have none);fixType, the row's kind, or nothing to take the row's;target:{"kind": "new_page"}for a page your store doesn't have yet, such as an about, policy, FAQ or comparison page. NameIO checks that the address has no page;- the text: the page's title, meta description, one new section and up to 8 questions and answers, and why;
storeFacts, for a service or trust worry or serious words: shipping and return facts from your policy page (days to return, who pays for returns, return countries, shipping countries, fewest and most days to deliver), each with the page's own words as itsquote. NameIO uses none of them until you tick them on the review page. NameIO then sets what the fix should move from the row, such as how many answers name you or how many have a price worry, and keeps where it stood on that check as "when you started". The assistant never sends that number:targetMetricandbaselineare refused, with the reason. AsignalIdthat a newer check no longer lists is refused, with the current rows. The page must be on your project's domain. There is one open draft per page, and up to 50 open drafts per project. To change a draft, ask the assistant to revise it. A part the revision leaves out stays as it is, and a part it sets to null is removed. A list of questions and answers replaces the draft's list, in its order, so a question the list leaves out is removed. NameIO tells the assistant what the revision removed, so it can tell you.
- Review it. Open Fixes in the sidebar, under AI Visibility. Each fix shows its kind, such as Price worry. The review page shows, top to bottom:
- For a fix saved for a row: its kind and what NameIO watches, such as "NameIO will watch: Named · 0 of 2 answers when you started". A new page says "This page doesn't exist yet. Create it at /pages/returns with this text." For serious words, What NameIO can't change names the sites the answers took the words from and the words never to repeat.
- Check before you use this text: every claim NameIO can't find on your page, such as prices, ratings and reviews, numbers, awards and certifications, superlatives, health and environmental claims, promises about returns, warranty or shipping, offers ("save 20%") and named competitors, and a title or description longer than search results show. A number is quoted with the words around it ("1 business day"). The same phrase in several fields is one item. Under What changes NameIO highlights each quoted phrase. Copy a request to change the text lists the claims, and any title or description to shorten, for your assistant.
- What changes: what the page says now next to the proposed text (a new page shows only the proposed text), with a copy button and a character count for each field. A new section or new questions say "Not on the page yet" under Now. Lines that start with "- " show as a bulleted list, as on your store.
- Why: the assistant's reasoning. NameIO doesn't check it.
- What NameIO saw: what it read on your page, and for each question the engines where you lost, who they named or linked to and which sites AI linked to, from that check, with its date.
- Store facts: each fact your assistant sent, with the policy page's words and, when NameIO found something to check, a line saying what. Tick the facts that are true. Only ticked facts go into the structured data, and only on a draft.
- Structured data (optional, built by NameIO): FAQ structured data from the questions and answers, and Product structured data for a product page that has none. Where your page already has it, NameIO says so instead. It never adds ratings or reviews, and adds a price only when it found one on your page. From the facts you ticked it adds your return policy (inside the product data it builds, or otherwise as your store's details) and, inside product data it builds, where you ship and how long delivery takes. Your assistant can't send structured data. On Shopify the code is wrapped so it runs on this page only; on other platforms it's a plain
<script type="application/ld+json">block. Its code and steps start closed under Show the code and steps. - How to apply it in your store's platform: the steps for each part, with copy buttons. NameIO uses the platform you pick under Store platform in Project settings. With Detect automatically, it uses the platform it found on the page it read for a fix, then the one your latest site check found, and when neither found one, a store with a Shopify connection counts as Shopify. If NameIO doesn't know your platform, the steps don't name its menus. On Shopify, structured data goes in a Custom Liquid section and needs an Online Store 2.0 theme; on an older theme, skip that step.
- For a fix saved for a row: its kind and what NameIO watches, such as "NameIO will watch: Named · 0 of 2 answers when you started". A new page says "This page doesn't exist yet. Create it at /pages/returns with this text." For serious words, What NameIO can't change names the sites the answers took the words from and the words never to repeat.
- Add it to your store yourself, using the steps. NameIO doesn't publish anything to your store. Then click Mark as added to my store, after ticking I checked each item under “Check before you use this text” if the fix has flags, or click Dismiss. Either can be undone with Back to draft. What the fix moved is under See what changed after a fix.
NameIO flags claims it can't find on your page. It looks for English wording, so on a page in another language it can miss claims. It doesn't verify claims, so one that isn't flagged may still be wrong. Read the text before you use it.
If your assistant isn't connected to NameIO, you can still get a draft:
- Click Write a fix with your AI assistant on a row under What to do next on your dashboard, or in the Competitor gap on Prompt Tracking. The page opens on Copy and paste. From a row, the row's tag and line are at the top, the instructions ask about its questions and the page field holds a suggested page. For serious words, a service or trust worry, or a missing policy page, you can choose A new page your store doesn't have yet and enter its address. Write a fix without connecting your AI assistant on the Fixes page does the same for up to 5 questions.
- Choose the page, click Get the instructions and paste them into any assistant. They hold the same brief, with the row's evidence and rules, and ask for the changes in a format NameIO can read.
- Paste the assistant's whole reply back into NameIO and save it. If something can't be used, such as HTML or a link to another site, NameIO lists every problem at once, one line each, so one corrected reply is enough. If the page already has a draft, NameIO asks before the reply replaces it, and the earlier draft is dismissed.
The draft opens on the same review page, marked Written from a pasted reply. To change it, paste the assistant's new reply under Paste the new reply. A part the reply leaves out stays as it is, and a list of questions and answers replaces the draft's list, in its order. If the reply removes a part, such as the page title or a question, NameIO names what it removes before it saves, and you choose Save anyway or Cancel.
See what changed after a fix
After you click Mark as added to my store, the fix's review page shows Before and after at the top, and your assistant can read the same numbers. Try: "Use NameIO to show what changed after my fix for the running-shoes page."
It shows what changed after you marked the fix as added to your store, not why. The comparison starts when you clicked Mark as added to my store, which may not be when the change went live. AI answers vary from check to check.
- What the fix should move: NameIO set it from the row the fix answers, such as how many answers have a price worry, how many name you, how many describe a competitor better, or whether a site check passes on your store. A fix started before NameIO kept the row watches "Named or linked", which AI answers below counts. The review page and your assistant lead with it, under What this fix should move: its name and what it counts, where it stood when you started (shown only; NameIO never compares against it), and the counts in the last check before you added the fix and in the Day 7 and Day 28 checks, counted again from your stored checks with the rules below. It then says where the count stands on the latest of those checks: Moved the right way when it changed by more than the metric's no-change band in the good direction (for worries, serious words and a competitor described better, fewer answers is the good direction), Moved the wrong way, No change, Too few answers (fewer than 10 answers on either side), or Waiting for the next check. More states say why nothing is compared: "Recounting earlier checks with the current word list" when NameIO's word lists changed between the two checks (the page leaves that check's count out until then), "The questions changed after you applied this fix" when no question and engine was answered in both, and "No check before you applied this fix, so nothing to compare". For a competitor described better it gives how many answers naming you both describe you better, the same and the competitor better, never a percent; the page shows them as a bar for each check, with the counts under it. For a site check it gives the status in words, such as "Failing on Oct 30, passing since Nov 3", and the page links to the check on Site readiness. It says which way the count moved, never that the fix did it. A results page you share shows the same words.
- On your page: when you mark the fix as added to your store, NameIO reads the page once and says whether it found the new title, description, section and questions and answers, with the date. If NameIO can't read your page, it says so. Check again reads the page again, once a minute at most.
- AI answers: for the fix's questions, "Named or linked in X of N answers" in three checks: the answers that name your store or link to your site. When some or all of the fix's questions have your store's name, a line under the count says that for those questions only links to your site count. The three checks:
- Before: your most recent check in the five weeks before you added the fix. With none, there's no Before.
- Day 7: the first check about a week after you added it.
- Day 28: the first check about four weeks after. Each shows its date, whether it was a scheduled check or Run now, and each competitor named, with how many of the answers named it ("Wrenfield was named in 2 of 3 answers"). Under the three cards, a short note gives the date and time each window opens, for example "Day 7 counts the first check that starts on or after Nov 7 at 6:23 PM." The windows open exactly 6 days and 27 days (24 hours each) after you marked the fix as added, and a check that starts before that time doesn't count. Next recheck gives the same opening time. When there were two or more checks before the fix, a line gives their counts in full, for example "Named or linked in 1 of 3 answers in each of the 2 checks before the fix was added (Oct 2 and Nov 2)." Once a Day 7 or Day 28 result is in, later checks don't change it. The page also lists every check since you added the fix, a few checks from before, and each question's result on each engine.
- Before: your most recent check in the five weeks before you added the fix. With none, there's no Before.
- What's counted: only answers both checks have, so the two sides compare like with like. For a question that has your store's name, only answers that link to your site count, and in the table a Named answer to it says "Not counted: only links count for this question." Checks in the first days after you add the fix are listed but not compared, and a result based on only a few answers is marked as a small sample.
- Recheck now: when the Day 7 or Day 28 check is due and no scheduled check will cover it, Recheck now opens the usual Run now dialog for the fix's questions, with its price. On the free plan, which has no scheduled checks, it's offered whenever a check is due. Before the next window opens, Next recheck gives its date and says who starts it: on the free plan, or with no scheduled check, you start it yourself; with a scheduled check, the first one on or after that time counts. Paused and archived questions are left out.
- Search Console: clicks, impressions and average position for the page, and when Google last crawled it, with whether that was before or after you marked the fix as added to your store.
- Google Analytics: sessions that started on the page, sessions from AI assistants, orders, and revenue in the property's currency. Both cover the 7 and 28 days before and after the day you marked the fix as added to your store, and that day is in neither side. Every number for the page has the whole store's number for the same days under it, so you can read a change on the page against a change everywhere. Each table shows its dates, the property it came from and its time zone: Search Console days are Pacific Time days, as Search Console reports them, and Analytics days are the property's own.
- When the numbers are ready: Google finishes a day's numbers after the day ends, so a comparison reads "ready from" a date until Google has reported its last day. Until Analytics has finished the last days, a note says the after side may still rise; if Search Console is late, a note says that side may be short.
- How to read them: a change says which way and by how much, such as "up 5" or "down 48 (12%)". Percentages appear only when both sides are large enough, and tables with few clicks or sessions are labeled Small numbers, because week-to-week swings can be that size. Average position is the same average Search Console shows, and a lower number is nearer the top of the results. Sessions from AI assistants count visits that arrive from ChatGPT, Perplexity, Gemini, Copilot, Claude and a few other assistants; many AI apps pass no referrer, so the count is a floor, not the total. Analytics marks numbers Approximate when Google withheld or estimated part of them.
- If Google isn't connected: each source stands alone. A source that isn't connected says so and offers Connect Search Console or Connect Analytics, which open Project settings → Integrations. An expired connection offers Reconnect, and if Google doesn't answer there's Try again. The AI answers and the page check show either way. See Google Search Console and Google Analytics.
Search Console and Analytics numbers come from your own Google connection when you open the page. Reading them uses no credits.
Moving the fix back to draft clears its before and after, and marking it as added again starts over. If another fix for the same page or questions was added after the Before check, the page says so: the checks after it reflect both.
In Fixes, the Results column shows where the checks stand for each fix added to your store. It reads "Day 7 from" a date and time until the Day 7 check opens, then Day 7 due or Day 7 checked, and the same for Day 28; under a Day 7 status it also says when Day 28 opens. A line under the table says that Day 7 and Day 28 open 6 and 27 days after you mark a fix as added, so a weekly check counts. On the free plan no check is scheduled, so the line says when results show instead ("Results show from Nov 7. Run a check that day to see them."). A draft or a dismissed fix shows a dash. A status of checked means a check ran in that window, which may have covered other questions; the review page has the counts for the fix's own questions. Once a Day 7 or Day 28 check is in (for a site check fix, once the site check has run since), the column also says where the fix's count stands: Moved the right way in green, Moved the wrong way in amber, and the other states, such as No change, in plain. A line above the table counts every fix you've added and where each stands, for example "6 applied: 2 moved the right way, 1 no change, 3 waiting". These are read on their own, so the list shows first and they follow a moment later, and are counted again at most every 10 minutes.
Your assistant reads the same before-and-after numbers as the review page, including Search Console and Analytics for the page and the whole store. Reading them uses no credits and doesn't start a check.
Share the results of a fix
Share results, at the top of Before and after, writes the results as a page for someone who isn't in NameIO. It asks first and tells you that anyone with the link can see the page.
Click Share and NameIO saves the page in Reports as "Fix results: /products/x (as of Oct 22, 2026)", with the type Fix results. Then it shows the page's /s/ link with a copy button.
The page holds what changed (the new text, and the page's old title and description when NameIO read them), the page check, the AI answers for each engine, the competitors those answers named, Search Console clicks, impressions and position, and Analytics sessions and sessions from AI assistants, with the whole store's beside the page's. It uses the review page's numbers and caveats, worded for a reader outside the store ("after the fix was marked as added to the store"). Its dates and times are in your time zone, as the review page shows them, and its footer names that time zone in plain words (for example, Central Time), or, for a time zone with no such name, gives its difference from UTC (for example, UTC-05:00). It leaves out orders, revenue and your assistant's reason for the fix.
The page is a copy as of the day you shared it, and it stays until you delete the report. Click Share results again to update the same page; the link stays the same while it's shared. To stop sharing, open the report in Reports and turn off Public link. Sharing again after that makes a new link.
Your assistant can also save and share a report of its own when you ask, but Share results is the way to send a fix's results.
Credits and plans
Lookups and checks use your organization's credits (1,000 credits are worth $1); reading saved data doesn't. Pro is $54 a month or $499 a year, plus tax where it applies. Both include 15,000 credits ($15) every month.
These use no credits:
- checking the connection, listing projects and creating a project
- project context and saved keywords
- rank tracker setup, tracker reads and cost estimates
- reports and report templates
- fix briefs, draft page fixes, the list of fixes and their before and after
- your assistant's tone reads: the answers to read and recording a read
- site audit lists, status, issues and pages, and deleting audits
- Search Console and Google Analytics reads
- AI visibility results, answers, sources, trends, counts over time and exports, and AI check cost estimates
- site readiness
These use credits:
- keyword research and keyword metrics
- Google search results and the sites that rank in them
- domain overview, domain keyword suggestions and ranked keywords
- backlinks overview and backlink profiles
- Prompt Explorer and Prompt Research
- AI visibility checks, one-off and scheduled
- rank checks, one-off and scheduled
Typical costs: keyword research, about 54 credits per keyword you look up; Google search results, about 5 credits per keyword; a domain overview or domain keyword suggestions, about 100 to 300 credits; a backlinks overview, about 50 credits for a domain. Ask your assistant for the cost before it runs a lookup.
Before an AI visibility check or a rank check, ask your assistant to show you the estimate. It sends the most the check may cost, and NameIO refuses the check if it now costs more. Your assistant can't buy credits or charge your card. If your organization runs out of credits, or asks for something your plan doesn't include, it can share a link to this section.
Prompt Explorer, Prompt Research, scheduled AI checks, rank checks, and tracking other engines need the Pro plan; the free plan tracks ChatGPT, Gemini and Google AI Overviews. Saving questions uses no credits on any plan. On the free plan, lookups and one-off AI checks also have a daily limit that all free accounts share, so it can run out early; it resets at midnight UTC (00:00 UTC), and credits you buy as a top-up aren't held to it.
Site audit crawling doesn't use credits. Lighthouse checks and the audit option Also check text that appears after the page opens do, and both need the Pro plan. On the free plan, an audit that asks for Lighthouse runs without it, and one that asks to check that text is refused. See Site audit.
Disconnect an assistant or revoke a key
To disconnect an assistant you connected through sign-in:
- Open Settings → Personal → Connected AI assistants.
- Open the app's row menu, click Disconnect and confirm.
The app loses access shortly after, usually within a minute. To use it again, connect it again from the app.
To revoke an API key:
- Open Settings → Personal → API keys.
- Open the key's row menu, click Revoke key and confirm.
Clients using the key stop working. API keys don't appear under Connected AI assistants.
Help and support
Email chris@getnameio.com from the address you sign in with. Say which assistant you use, for example Claude or ChatGPT, and what you asked it. Never send an API key or a password.
- The connection stopped working. Disconnect NameIO in your assistant and connect again. A connection you don't use for 30 days stops working until you connect it again.
- Something NameIO returned looks wrong or harmful, for example a wrong count, a broken link or an answer quoted in the wrong place. Email us the question you asked your assistant and roughly when. NameIO quotes AI answers as other assistants gave them, so a quoted answer can be wrong; tell us if NameIO showed it as its own finding.
- Your data. Your data: see, export, delete shows how to export your results, disconnect an assistant and delete your account.
- A security problem. Email chris@getnameio.com with "Security" in the subject and the steps to reproduce it. Please don't test against other people's accounts.
Technical details
For client and agent developers.
- Endpoint:
https://app.getnameio.com/mcp, Streamable HTTP over POST. - Discovery:
/.well-known/mcp.jsonon getnameio.com points to the server. - OAuth endpoints:
https://app.getnameio.com/api/auth/oauth2/authorize,https://app.getnameio.com/api/auth/oauth2/token, andhttps://app.getnameio.com/api/auth/oauth2/registerfor Dynamic Client Registration. Custom-scheme redirect URIs are allowed. Clients can also identify themselves with a Client ID Metadata Document. - Scopes:
mcp(required: reads and lookups),mcp:write(changes) andoffline_access. A client that names no scopes gets all three. - API keys: send
Authorization: Bearer nio_...orx-api-key: nio_.... If you send too many requests, the server returns 429 withRetry-After.
What's not done
- MCP clients that run inside a web page on another site can't connect.
- Cursor is the only client with a one-click install. You add the server in the others yourself.
- NameIO isn't in the Microsoft 365 Agent Store; in Microsoft 365 Copilot it's added through Copilot Studio.
- AI visibility doesn't track Claude.
- Publishing a fix straight to your Shopify store isn't available. You paste each fix in yourself.
- Search Console and Analytics numbers on a fix cover the page as a whole, not single queries or products. If you disconnect Google, they leave the review page; a page you shared keeps its copy.
- Site checks don't read Google's speed data from real visitors (the Chrome UX Report, which needs a Google API key), your Google Merchant Center or Microsoft Merchant Center account, or your images, so image sizes are checked only from a connected Shopify store.
- The only store connection is Shopify. Stores on other platforms give their product data through the public product list or a product feed.
- NameIO sends no alerts or emails when a count changes.
- You can't dismiss a row of What to do next.