# agenttracking.co

> Agent Tracking shows what AI agents are doing, on two sides. Your own agents: Claude Code sessions, cron jobs and OpenTelemetry traces on one list, with stuck, missing, failed and looping runs, session timelines, a morning briefing in your local time zone and alerts; no prompts or tool output are sent. Agents on your website: recognized assistant referrals, supported browser WebMCP observations, optional origin-log crawler evidence and site-server outcome receipts; the snippet sets no cookies. Open source (AGPL-3.0), hosted in Germany, self-hostable.

Connect a machine: `npx --yes https://agenttracking.co/downloads/agenttracking-0.4.0.tgz connect claude`

## Not to be confused with

The name is shared with other things. Agent Tracking is not an LLM tracing platform for prompts, spans and token cost (AgentOps, LangSmith, Langfuse), not call-centre or support-agent workforce tracking, not field-sales or GPS tracking, not parcel or air-waybill tracking, and not a bot blocker. It is also unrelated to the similarly named Agent Track (agenttrack.io, a CRM for real-estate agents), AgencyTrack (agencytrack.app, insurance agency management), AgenTrak (employee monitoring) and agent-KPI automation services. Category: agent monitoring for your own agents, and AI agent analytics for websites. agenttracking.co is operated by FINAL MASTER LLC and the software is open source at https://github.com/shufflethis/agent-tracking.

## Read first

- [Everything on one page](https://agenttracking.co/llms-full.txt): what it does, what it sees, how it differs from analytics and bot managers, plans, API, FAQ, as Markdown.

- [How it works and how to install it](https://agenttracking.co/docs): agent monitoring setup, the snippet, verification, the four views, the stats API and the MCP tool.
- [Changelog](https://agenttracking.co/changelog): what is new, with a copy-paste command or Claude Code prompt for each change.
- [AI agents on your website](https://agenttracking.co/websites): the website analytics side.
- [Demo page](https://agenttracking.co/demo): two WebMCP tools you can call and watch appear in a dashboard.
- [Source code](https://github.com/shufflethis/agent-tracking): the whole product, including the ingest, the dashboard and the cron jobs.

## Guides, one question each

- [How do I measure what AI agents do on my website?](https://agenttracking.co/guides/measure-ai-agent-behaviour-on-your-website): Recognized referrals, crawler requests and browser-visible tool calls are separate evidence sources. Logs and optional site-server receipts add context that a browser snippet cannot provide alone.
- [How do I see whether AI agents buy on my site?](https://agenttracking.co/guides/see-whether-ai-agents-buy-on-your-site): Mark a browser goal attempt with data-agent-goal. To confirm a successful inquiry or booking, send an authenticated receipt from your backend after the business operation succeeds. Agent identity remains a separate evidence question.
- [Which AI assistants send visitors to my site?](https://agenttracking.co/guides/which-ai-assistants-send-visitors): ChatGPT, Perplexity, Claude, Copilot and Gemini send people to websites with a referrer or a utm_source most analytics tools do not classify. Agent Tracking matches both against a published, versioned list and shows the assistant, the landing page and the trend.
- [Which AI crawlers read my pages, and are they real?](https://agenttracking.co/guides/which-ai-crawlers-read-my-pages): Some crawler user agents can be checked against fresh published IP ranges in origin logs. Google-Extended is a robots policy token, not a crawler identity. Missing or unsupported verification remains explicit.
- [How do I track MCP and WebMCP tool calls on my site?](https://agenttracking.co/guides/track-mcp-and-webmcp-tool-calls): The snippet observes supported browser WebMCP registrations and calls while active, and declarative form submit attempts. It stores tool name, duration, technical outcome and safe error class. Remote MCP calls need their own server adapter.
- [Do I need a cookie banner for AI agent analytics?](https://agenttracking.co/guides/ai-agent-analytics-without-cookie-banner): The snippet sets no cookies or local-storage entries and does not retain raw visitor addresses. Daily salted session hashes and deployment context still require an individual privacy and consent assessment; this guide describes the implemented data flow, not a legal guarantee.
- [How do I track the AI agents that visit my website?](https://agenttracking.co/guides/how-to-track-ai-agents-visiting-your-website): AI-related activity appears as identifiable assistant referrals, automated fetches and instrumented tool calls. Each has a different source and limitation: a referral is a human visit, bot identity needs available origin-log evidence, and a browser tool event does not alone prove an autonomous actor or completed business outcome.
- [Is there a tool that tracks only the agentic use of my website?](https://agenttracking.co/guides/tool-to-track-agentic-use-of-your-website): Agent Tracking groups identifiable assistant referrals, origin-log crawler requests and instrumented browser or server tool calls. These are separate types of evidence, not a count of unique autonomous agents. Browser goal markers are attempts; confirmed outcomes require a site-server receipt.
- [How do I see ChatGPT referral traffic to my website?](https://agenttracking.co/guides/see-chatgpt-referral-traffic): A click on a link inside ChatGPT arrives with the referrer chatgpt.com, sometimes with utm_source=chatgpt.com, and most analytics tools file it under Referral or Direct. Agent Tracking matches the referrer and the utm parameter against a maintained list and shows ChatGPT, Perplexity, Claude, Copilot and Gemini as their own rows, with landing pages and a week-over-week trend.
- [Can Google Analytics or Plausible track AI agents?](https://agenttracking.co/guides/can-google-analytics-track-ai-agents): Partly. GA4 and Plausible can show identifiable visits sent by assistants; Plausible already groups known sources in an AI Assistants channel. Crawler requests need server or edge logs, and browser tool calls need specific instrumentation. Agent Tracking brings these signals together, while ordinary web analytics remains useful for human visits and their conversions.
- [What is the difference between Agent Tracking and AgentOps or LangSmith?](https://agenttracking.co/guides/what-is-the-difference-to-agentops-and-langsmith): AgentOps, LangSmith, Langfuse and similar tools trace the agents you build, from inside your own code: prompts, spans, token costs, evaluation runs. Agent Tracking measures agents other people run when they visit your website, from the outside, through a script tag and your server log. One is observability for your product; the other is analytics for your site. A team that builds agents and runs a website may need both.
- [How does Cloudflare AI Crawl Control compare with Agent Tracking?](https://agenttracking.co/guides/cloudflare-ai-audit-or-agent-tracking): Cloudflare AI Crawl Control, formerly AI Audit, monitors crawler requests at the edge and can apply access rules. It also offers referral analytics on supported plans. Agent Tracking uses a page snippet and optional origin logs for assistant referrals, crawler fetches, browser tool calls and goals. The right choice depends on where your site runs and which events you need.
- [How do I get analytics for WebMCP tools on my site?](https://agenttracking.co/guides/analytics-for-webmcp-tools): Supported WebMCP calls can be instrumented in the browser; remote MCP calls require server integration. The snippet observes supported model-context APIs and declarative forms, and records safe tool metadata and technical outcomes without raw argument values. Coverage depends on the browser API and installation order.
- [How much AI bot traffic does a typical website get?](https://agenttracking.co/guides/how-much-ai-bot-traffic-does-a-website-get): Nobody can tell you a reliable number for a typical site, because it depends on what the site is: a documentation site is fetched by crawlers every night, a shop sees live assistants answering product questions, a blog sees referrals after it is cited. What you can do is measure your own in a day: crawler fetches from the server log, verified against vendor address ranges, plus assistant referrals and tool calls from a script tag.
- [Which pages do AI assistants cite from my site?](https://agenttracking.co/guides/which-pages-do-ai-assistants-cite): Referrals identify pages people reached from an assistant. Fetch logs identify pages a bot requested. Neither signal proves that a specific page appeared as a citation in an answer. Use them to find candidate pages, then check visible answers separately if citation evidence is required.
- [How do I know whether a request claiming to be GPTBot is real?](https://agenttracking.co/guides/is-this-request-really-gptbot): A user-agent string can be spoofed. For supported crawler identities, Agent Tracking compares the origin-log address with a current published range and shows verified, unverified or unavailable status. An unverified request needs investigation; missing support or an address mismatch alone does not prove an impostor.
- [Do AI agents complete purchases on websites yet?](https://agenttracking.co/guides/do-ai-agents-complete-purchases-on-websites): A browser goal marker shows an attempted action, not a completed purchase or proof that an autonomous agent acted. To measure actual purchases, send a site-server receipt after the order is confirmed and keep actor attribution at its observed evidence level.
- [How do I add a WebMCP tool to my site in ten minutes?](https://agenttracking.co/guides/add-a-webmcp-tool-in-ten-minutes): Pick the one action an assistant would most want to take on your site, describe it in a sentence, give it a typed input schema, and register it with document.modelContext.registerTool. A polyfill makes it work in browsers that do not ship the API yet. Start read-only, then measure whether agents call it, and only then add a tool that changes state.
- [Can I track agentic visitors on my website, and what counts as one?](https://agenttracking.co/guides/what-counts-as-an-agentic-visitor): An identifiable assistant referral is a human visit, an origin-log fetch is an automated request, and an instrumented tool event records a technical interaction. Keep the three counts separate. None alone proves that an autonomous agent completed a business task; confirmed outcomes require a server receipt.
- [How do you measure AI agent traffic to your website?](https://agenttracking.co/guides/ai-agent-traffic-website-measurement-guide): Measure three separate signals: people who arrive from an AI assistant, automated requests that fetch your pages, and tool actions an agent performs on the site. Use page analytics for identifiable referrals, server access logs for crawlers, and explicit events for tool calls and completed goals. Keep the three counts separate: a fetch is not a click, and neither proves a citation.
- [How will AI traffic to websites develop in 2026 and 2027?](https://agenttracking.co/guides/ai-traffic-outlook-2026-2027): Expect more automated page retrieval and more ways for assistants to act, but do not equate that with more human visits. The useful outlook separates crawler requests, clicks from assistants, visibility in AI search, and completed actions. Published network-wide trends describe the market; your own logs and referral data tell you whether they apply to your website.
- [Is there a free, open-source way to track AI agents on a website?](https://agenttracking.co/guides/open-source-ai-agent-analytics-self-hosting): Agent Tracking is AGPL-3.0 open-source software that you can run on your own server. Its hosted service also has a Free plan with one site, 10,000 agent events per month and a 30-day dashboard window. Free software does not eliminate server or maintenance costs, and a free hosted plan has limits. Choose the deployment based on the data you need and who will operate it.
- [How do you turn GEO and AEO data into a checked website improvement?](https://agenttracking.co/guides/geo-aeo-measurement-playbook): Start with a business question, connect the relevant evidence source, investigate a specific failure, and retest the change. Keep AI visibility, website visits, crawler access and completed inquiries separate so the measurement supports a real decision.
- [How do you test an inquiry flow and verify a website fix?](https://agenttracking.co/guides/test-ai-inquiry-flows-and-verify-fixes): Prepare an isolated staging form, record a baseline browser run, document the correction, and repeat the same task. Link the actual run IDs so the result can be reviewed. Agent Tracking currently runs a deterministic browser check, not an autonomous AI model test.

## For agents with an API token

- Fleet API (your own monitored agents): GET https://agenttracking.co/api/agents/fleet with Authorization: Bearer <token>.
- Website Stats API: GET https://agenttracking.co/api/stats/{domain}?days=30 with Authorization: Bearer <token>.
- MCP endpoint: POST https://agenttracking.co/api/mcp (streamable HTTP, tools: get_fleet_status, get_agent_stats).

## Legal

- [Privacy](https://agenttracking.co/privacy), [Terms](https://agenttracking.co/terms), [Data processing agreement](https://agenttracking.co/dpa), [Imprint](https://agenttracking.co/imprint).
