You've likely asked yourself these questions recently:
Which AI crawlers visited our website last week?
Which product pages is ChatGPT crawling most?
Did bot traffic increase after our latest release?
Which crawlers are generating the most 404s?
Where is our suspicious traffic coming from?
Until now, answering questions like these usually meant opening an analytics dashboard, selecting the right filters, comparing date ranges and digging through tables.
HoneyLog now gives you another option.
With our new Model Context Protocol (MCP) server, you can connect HoneyLog directly to ChatGPT, Claude, Cursor, VS Code or another MCP-compatible AI assistant and query your server-side traffic data using natural language.
Connect it once, then start asking questions.
From dashboards to conversations
HoneyLog analyzes traffic at the server or CDN level, where requests are visible before any browser-side JavaScript is executed.
That allows HoneyLog to distinguish between human visitors, verified crawlers, suspicious automation and malicious bots. This includes AI agents and search engine crawlers that traditional client-side analytics cannot detect.
The dashboard is still the best place to monitor that traffic visually. But sometimes what you want isn't another dashboard, but an clear answer.
Which AI bots crawled our new product category since it launched?
Compare Googlebot traffic this month with last month and show me the pages with the biggest decline.
Which bots caused yesterday's traffic spike?
Show me the countries generating the most malicious bot requests.
That is where MCP comes in.
Instead of manually translating your question into dimensions, filters and date ranges, your AI assistant can do that work for you and return the result in the format you need.
What is MCP?
MCP, or the Model Context Protocol, is an open standard that allows AI assistants to connect to external tools and data sources.
Think of it as a common language between an AI assistant and the software you already use.
Once HoneyLog is connected, the assistant can determine which sites it has access to, understand the metrics and dimensions available for them, query HoneyLog and use the returned data to answer your question.
You don't have to know the API, you won't need to write a query, or export a CSV and upload it into a chat every time you want to investigate something. Instead, you can simply ask.
What can your AI assistant actually access?
HoneyLog's MCP server uses the same underlying traffic-reporting API as the HoneyLog API itself.
Through MCP, an assistant can analyze traffic over time and break it down using dimensions including:
- Pages and URLs
- Individual bots
- Bot categories
- Human, verified, suspicious and malicious traffic
- Countries
- Referrers
- Browsers and operating systems
- HTTP status codes
- Sites within a network
It can also work with metrics such as requests, users and sessions, response times, bytes sent, 2xx requests, redirects, 4xx responses and server errors.
That opens up considerably more than simple questions like “how much bot traffic did we get?”
7 things you can do with HoneyLog MCP
1. Investigate your AI visibility
AI search is creating an entirely new type of website consumption.
OpenAI, Anthropic, Perplexity, Google and other companies operate different crawlers for search, retrieval, indexing and model-related purposes.
HoneyLog lets you see that activity directly at the server level. MCP makes exploring it conversational.
Try asking:
Which AI bots crawled our website during the last 30 days? Show their request volume and the five pages each bot accessed most.
Or:
Which pages under /blog/ received the biggest increase in AI crawler traffic compared with the previous month?
Instead of simply knowing that AI crawlers exist, you can start understanding what they actually want from your website.
2. Find content that search and AI crawlers are ignoring
Publishing a page does not mean machines are actually discovering it.
For SEO and GEO teams, server-side crawl activity provides another signal for understanding how search engines and AI platforms interact with your content.
Ask:
Which pages published in this section received little or no traffic from Googlebot this month?
Or:
Which pages are heavily crawled by Google but rarely accessed by AI bots?
Questions like these can help reveal differences between traditional search discovery and emerging AI discovery.
3. Diagnose crawl problems without digging through logs
Server logs can tell you when crawlers encounter redirects, broken pages, slow responses or server errors.
But raw log analysis is rarely pleasant.
With MCP, you can ask questions directly:
Which URLs generated the most 404 responses for verified crawlers during the last 30 days?
Show me the pages with the highest average response time when accessed by Googlebot.
Did 5xx errors for bots increase after our latest deployment?
The assistant can query the relevant traffic data, compare periods and summarize what changed.
For SEO, engineering and infrastructure teams, that can turn a lengthy investigation into a much faster first diagnostic.
4. Understand sudden bot traffic spikes
Not every traffic spike is a success.
Sometimes it is Google recrawling thousands of URLs.
Sometimes it is an AI crawler suddenly consuming an entire section of your site.
Sometimes it is an unidentified scraper.
And sometimes it is malicious automation.
Ask HoneyLog:
Bot traffic jumped yesterday. Which bots were responsible, and which pages did they request?
Or:
Compare suspicious bot traffic this week with the previous week and identify the largest changes.
The important part is not only seeing that traffic increased. It is understanding what produced the increase.
5. See what AI systems care about on an e-commerce site
For e-commerce teams, AI crawler activity introduces a new layer of product visibility.
Which categories are AI systems exploring?
Which products are being repeatedly fetched?
Are crawlers concentrating on product detail pages, category pages or informational content?
For example:
Which product URLs received the most requests from AI crawlers in the last 30 days?
Or:
Compare AI crawler activity across our /shoes/, /bags/ and /accessories/ sections.
This doesn't tell you by itself why an AI system will recommend a particular product.
But it gives you something much more concrete than speculation: what AI systems are actually requesting from your website.
6. Investigate suspicious and malicious traffic
HoneyLog does more than identify known crawlers.
It also separates traffic that is clearly automated but cannot be reliably attributed from traffic identified as malicious.
With MCP, an investigation can begin with questions such as:
Which countries generated the most malicious bot requests this week?
Which pages are targeted most often by suspicious bots?
Show me malicious traffic over the last 30 days and highlight unusual spikes.
That makes server-side traffic intelligence accessible even when the person investigating the issue does not regularly work with raw logs.
7. Analyze several websites at once
HoneyLog Networks let organizations group multiple properties together.
With the appropriate permissions, MCP can query network-level data too.
That is useful for agencies, media groups, marketplaces or companies operating multiple domains.
You could ask:
Which of our websites received the most AI bot traffic this month?
Give me a network-wide summary of verified, suspicious and malicious bot traffic for the last seven days.
Which sites experienced the largest increase in AI crawler activity compared with the previous month?
Instead of opening every property individually, you can start from the question and drill down into the sites that need attention.
Your data stays read-only
Giving an AI assistant access to analytics data naturally raises an important question:
What can it change?
Nothing.
HoneyLog's MCP implementation is strictly read-only.
The assistant can query the traffic data your token permits it to access, but it cannot change your HoneyLog configuration, delete information or modify your billing.
Access is controlled through HoneyLog team API tokens.
You can grant access to sites and, when needed, networks. If the token is deleted, its access stops.
MCP also uses the same permissions and API limits as the HoneyLog API rather than creating a separate unrestricted path to your data.
MCP doesn't replace the HoneyLog dashboard
Conversational analytics and dashboards solve different problems.
The HoneyLog dashboard remains the easiest way to visually explore traffic, monitor trends, work with features such as Visibility Radar, Sitemap analysis, Bot Conversions and robots.txt intelligence, and see what is happening across your properties.
MCP is particularly useful when you already know the question you want answered.
The dashboard is exploration-first.
MCP is question-first.
And because your AI assistant can continue the conversation, an initial question can quickly become an investigation:
Which AI bots grew the most this month?
Then:
Show me only OpenAI and Anthropic.
Then:
Which pages explain the increase?
Then:
Compare those pages with the previous month.
Then:
Summarize the findings for our SEO team.
The underlying data stays the same. What changes is how quickly you can move through it.
How to connect HoneyLog to your AI assistant
Setup only takes a few steps.
- Open Settings → API tokens in HoneyLog.
- Create a team token with the appropriate read permissions.
- Add the HoneyLog MCP server to your MCP-compatible client.
- Give the client your token.
- Start asking questions.
HoneyLog currently works with MCP-compatible tools including ChatGPT, Claude, Cursor and VS Code, with the same endpoint working across your permitted sites, custom views and networks.
For configuration examples and detailed setup instructions, see the HoneyLog MCP documentation.
Analytics shouldn't stop at the dashboard
Server-side analytics gives you access to a part of the web that conventional browser analytics cannot fully describe: crawlers, AI agents, scrapers and automated traffic interacting directly with your infrastructure.
HoneyLog already makes that traffic visible.
MCP makes it conversational.
Instead of adapting your question to an analytics interface, you can now bring the data into the AI tools where you already research, code, investigate and work.
Connect HoneyLog to your AI assistant and ask your first question.
Try HoneyLog
New to HoneyLog?
Start a 14-day free trial, connect your website and see who — and what — is actually visiting it.
No credit card required.