AnythingLLM + Answer Socrates
Add Answer Socrates to AnythingLLM and do your SEO research with its agent. Ask in plain English and AnythingLLM pulls real Google data: hundreds of real searches for any keyword, 12 months of Google Trends and the searches rising right now. Then it groups the lot into a content plan.
Last updated: October 2026
Add Answer Socrates to AnythingLLM as a remote MCP server: one entry in anythingllm_mcp_servers.json and an API key from your Answer Socrates account. The desktop app needs v1.8.0 or later. On a self-hosted Docker install, whoever runs the server adds it.
Create an API key on your Answer Socrates Account page. It is shown once, so copy it somewhere safe.
Open the Agent Skills page once so AnythingLLM creates anythingllm_mcp_servers.json, then open that file in the plugins folder of your AnythingLLM storage directory.
Paste this in, with your own key in place of as_live_YOUR_KEY, and save. Keep "type": "streamable": without it AnythingLLM assumes the older SSE transport.
{
"mcpServers": {
"answersocrates": {
"type": "streamable",
"url": "https://answersocrates.com/api/mcp",
"headers": {
"Authorization": "Bearer as_live_YOUR_KEY"
}
}
}
}Back on Agent Skills, click Refresh. Check answersocrates is listed under your MCP servers with no error. No restart needed.
Start your message with @agent so AnythingLLM can use its tools, then ask in plain English, for example: @agent use Answer Socrates to pull the keywords for camping gear.
This file holds your key in plain text. On a shared Docker install, everyone using the agent draws on that one Answer Socrates account. Small local models often skip tool calls, so pick a model that handles tools well.
Ask AnythingLLM for keywords on its own and your model guesses from what it learned in training. With Answer Socrates connected the agent pulls what people actually type into Google: hundreds of real search phrases per keyword, sorted into four groups.
How, what, why, can: the questions people ask about your topic, ready for FAQs and blog posts.
For, with, without, near: the qualifiers that show what people actually want.
Vs, or, like, and: the head-to-head searches that sit right before a buying decision.
The full a to z list of Google suggestions for your keyword.
What a keyword pull returns, from a real Answer Socrates session:
get_keywordssave_to_project
1,187 keywords for creatine gummies: 160 Questions (“are creatine gummies as effective as powder” asked five different ways: pure format skepticism), 166 Comparisons, 81 Prepositions, 30 time-based and 750 autocomplete phrases.
Saved to Supplements. Your team can see the full list in the app now. Want them as a CSV too? That's free.
Planning videos? Ask for the YouTube version and AnythingLLM pulls what people type into YouTube's own search box, sorted into the same four groups. YouTube searches lean towards tutorials, reviews and how-tos, so the list is different from Google's.
get_youtube_keywords
Here are the YouTube searches for ultralight backpacking gear, grouped into Questions, Prepositions, Comparisons and Autocomplete, with the how-to questions up top. Want the Google version too?
Name a market and the AnythingLLM agent maps it: what is climbing, what is seasonal and what is fading, from 12 months of Google Trends data in any country. Rising searches adds the related searches growing fastest right now.
Weekly Google Trends interest for any keyword across the last 12 months, in any country.
See when a keyword peaks and when it dips, so your content is live before the rush.
Run the same keyword in the UK, the US or any other country and chart them side by side.
Compare 10 or more keywords at once to see which are climbing and which are fading.
The related searches growing fastest right now, including brand-new Breakouts.
Ask AnythingLLM to chart it, table it or export it. Reworking data it already pulled costs nothing.
And much more. Ask in plain English and AnythingLLM picks the right tools, like this comparison of one keyword in two countries:
get_rising_topicsget_trends ×12
Over the last 12 months, the big movers are clear whey +190%, creatine gummies +94% (a real format shift, not hype), electrolytes +63% and magnesium glycinate +62%. Classic whey is resurging (+64%) while ashwagandha is the only decliner (−5%). That trend has peaked.
In the rising searches, Nutricost whey is up +1,850% and two brands hit Breakout. Almost everything peaks in April, so publish by March to catch the crest. Want me to pull the full keyword set for creatine gummies? (1 request)
Use the list AnythingLLM just pulled, or paste in a list of your own. Answer Socrates groups it into topics and labels every keyword with search intent and funnel stage, so you know what to write first. Clustering is paid in cluster credits, 1 per keyword, not requests.
cluster_keywords
931 keywords into 18 topics. The biggest is what keyword research is at 206 keywords, almost all top-of-funnel. The money is elsewhere: free keyword research tools (144) and paid tools and pricing (72) are both commercial and mid-funnel.
Only 13% didn't fit a topic. Want me to list any cluster in full, or export the lot as a CSV?
AnythingLLM is great at chatting with your own documents, privately, with the model of your choice. What it lacks is live search data, and that is the part Answer Socrates adds.
| Task | AnythingLLM alone | With Answer Socrates |
|---|---|---|
| Keyword ideas | Guesses from training data | Hundreds of real Google searches per keyword |
| Questions people ask | Plausible-sounding questions | Real questions from Google autocomplete |
| Search trends | No Google Trends data | 12 months of Google Trends, any country |
| What's rising now | No live Google search data | Google's rising searches, Breakouts included |
| Clustering | Groups whatever you paste | Topics, intent and funnel stage for every keyword |
| Saving the work | Stays in the chat | Saved to your projects for the whole team |
Anything you would ask an SEO on your team. Start with @agent, name the topic, the market and what you want back, and AnythingLLM can lay the results out as tables you can copy.
Keyword research
“@agent use Answer Socrates to pull the keywords for camping gear and show me the questions first.”
Content plan
“@agent find what's rising in camping gear, pull keywords for the top three topics and cluster them into a content plan.”
Trends
“@agent compare interest in camping tents, sleeping bags and camping stoves over the last 12 months. Which one is growing?”
Your website
“@agent here is the copy from my homepage. Do keyword research for the topics it covers.”
Search intent
“@agent group these camping gear keywords by topic and tell me which ones are commercial.”
Export
“@agent cluster the questions and give me the plan as a CSV I can paste into a spreadsheet.”
That is a different job, on the same account. This page is about using AnythingLLM to do your SEO. To find out whether AI assistants recommend your brand, the LLM Brand Tracker runs your prompts across ChatGPT, Claude, Perplexity, Gemini and more, and shows who gets mentioned.
Track your brand in AI searchConnecting is free. The data comes from your Answer Socrates plan, where every plan includes 3 requests a month to try it and Seneca and above include the full allowances. Your plan has two separate allowances:
1 request per data call: rising topics, Google Trends or a full keyword pull. Checking your usage, saving to projects and exports are free.
1 credit per keyword you cluster, 10 to 1,000 keywords per call. These are the same cluster credits as the clustering tool in the app, and clustering never spends requests.
| Plan | Requests | Cluster credits |
|---|---|---|
| Free | 3 /mo | ✕ |
| Lite | 3 /mo | ✕ |
| Seneca | 500 /mo | 12,000 /mo |
| Aurelius | 2,500 /mo | 40,000 /mo |
| Senate | 25,000 /mo | 100,000 /mo |
Free and Lite include 3 requests a month to try it, renewed on the 1st, with no clustering through MCP. Yearly plans get 12 months of both allowances up front. Team plans share one pool across all seats.
See plansSame account, same tools. Answer Socrates works in any assistant that supports MCP.