# Agent Analytics Blog

> Updates, guides, and thoughts on analytics for the agent era.

## Blog Posts

- [Agent Analytics Now Supports WebMCP](https://blog.agentanalytics.sh/blog/agent-analytics-webmcp/): Open a live analytics page, ask a product question, and let your AI agent call structured read-only tools while you inspect the same evidence on screen.
- [Agent Analytics Now Supports auth.md](https://blog.agentanalytics.sh/blog/auth-md-agent-analytics/): Agent Analytics now publishes auth.md discovery and turns human browser approval into a scoped, revocable agent session.
- [Best Product Analytics for AI Builders](https://blog.agentanalytics.sh/blog/best-product-analytics-for-ai-builders/): Compare Agent Analytics, PostHog, Mixpanel, Amplitude, Plausible, Umami, and Google Analytics for teams that need product analytics their AI agents can use.
- [Hermes Skill Bundles Make Agent Workflows Feel Native](https://blog.agentanalytics.sh/blog/hermes-skill-bundles-agent-analytics/): Hermes just introduced skill bundles. They turn repeatable agent workflows into one slash command, so users can name a job once and let Hermes load the right skills and instructions every time.
- [Jobs To Be Done for AI Landing Pages](https://blog.agentanalytics.sh/blog/jobs-to-be-done-ai-landing-pages/): AI agents can mine customer language, but humans still choose the job and success signal. Use Jobs To Be Done to turn AI landing pages into measured job hypotheses.
- [Teach Your AI Agent AIDA for Landing Page Analytics](https://blog.agentanalytics.sh/blog/aida-for-ai-agent-landing-page-analytics/): AIDA is old. The new workflow is asking your AI agent to map Attention, Interest, Desire, and Action to live landing-page analytics before it writes variants.
- [How to Use STP Marketing When AI Can Write for Everyone](https://blog.agentanalytics.sh/blog/stp-marketing-ai-builders/): Segmentation, targeting, and positioning matter more when your AI agent can create endless pages. Use STP to force one segment, one target, one promise, and one measurable activation loop.
- [YC Is Right About AI-Native Startups. They Need Closed-Loop Analytics.](https://blog.agentanalytics.sh/blog/ai-native-companies-need-ai-native-first-analytics/): YC's AI-native company advice points to one practical operating model: measurable surfaces, project context, portfolio context, and analytics your existing AI agents can use when deciding what to do next.
- [You Can Now Scale Depth as Well as Width With Hermes Subagents. Agent Analytics Got 140% More Growth Bets](https://blog.agentanalytics.sh/blog/hermes-subagent-delegation-agent-analytics/): Hermes just released Subagent Delegation. Agent Analytics used that exact model in its skill for growth audits, and our dogfood run produced 140% more growth bets, 140% more evidence references, and 125% higher output density.
- [If You Use Hermes to Handle Your Projects, You Need Agent-Readable Web Analytics](https://blog.agentanalytics.sh/blog/hermes-needs-agent-readable-web-analytics/): Hermes can handle your projects, ship changes, and keep context. Agent-readable analytics closes the loop so Hermes can see what users actually did after the work shipped.
- [If You Build for AI Agents, Your Upgrade Flow Needs to Change](https://blog.agentanalytics.sh/blog/agent-friendly-upgrade-links/): When an AI agent uses your product, the upgrade moment should carry the task context, the paid capability, and a safe human approval handoff.
- [Your Product Isn’t One Website Anymore](https://blog.agentanalytics.sh/blog/project-context-for-ai-agents/): A practical model for agent-readable analytics: portfolios connect related projects, projects keep local product truth, and surfaces are where users encounter each project.
- [A Product Growth Scanner for Your Website](https://blog.agentanalytics.sh/blog/product-growth-scanner/): Agent Analytics turns your public site into a prioritized measurement plan and shows your coding agent the data you are not collecting yet.
- [Autoresearch Growth Loops Need Reality Checks](https://blog.agentanalytics.sh/blog/autoresearch-growth-loop-agent-analytics/): LLM judges are useful for generating pressure, but they drift when they judge themselves. The better loop ships variants, waits for real behavior, and starts the next round from evidence.
- [The Feedback Endpoint Claude Code and Codex Need](https://blog.agentanalytics.sh/blog/feedback-endpoint-ai-agents-need/): If Claude Code, Codex, OpenClaw, Paperclip, or another AI agent uses your product, feedback should not be a UI form. Give the agent a documented endpoint.
- [The Login Pattern AI Agents Need](https://blog.agentanalytics.sh/blog/login-pattern-ai-agents-need/): A safe login pattern for AI agents: the human owns identity, the agent owns the work, and no one pastes API keys into chat.
- [Session Paths: Entry Pages, Exit Pages, and What Happens Between](https://blog.agentanalytics.sh/blog/session-paths-entry-exit-pages-for-ai-agents/): Session paths make entry pages, exit pages, and the steps between agent-readable, so your agent knows what to investigate next.
- [If You Use 🗄️ Cabinet, You Need Agent-Readable Web Analytics](https://blog.agentanalytics.sh/blog/cabinet-teams-need-end-user-analytics/): 🗄️ Cabinet gives you the knowledge base and AI team. Agent-readable analytics gives your Data Analyst real user outcomes to measure and improve.
- [Best Analytics for AI-Built Side Projects](https://blog.agentanalytics.sh/blog/best-analytics-for-ai-built-side-projects/): A practical comparison of Plausible, Umami, PostHog, Mixpanel, Google Analytics, and Agent Analytics for AI-built side projects and agent-led workflows.
- [If You Use 📎Paperclip, You Need Agent-Readable Web Analytics](https://blog.agentanalytics.sh/blog/paperclip-companies-need-agent-readable-analytics/): 📎Paperclip orchestrates zero-human companies. Agent-readable analytics shows whether users reached install, signup, API keys, and first project.
- [Vibe Coding Made Me Ship More Projects Than I Could Track Manually](https://blog.agentanalytics.sh/blog/vibe-coding-made-me-ship-more-projects-than-i-could-track-manually/): AI coding tools made it easy to launch more side projects than dashboard workflows can track. Agent Analytics gives your AI agent one measurement layer.
- [Why We Built Live Monitor for macOS](https://blog.agentanalytics.sh/blog/why-we-built-live-monitor-for-macos/): A faster way to watch live traffic, QA trackers, and keep real-time context available for you and the agents helping you ship.
- [AARRR for Solo Builders: Let Your AI Agent Run Your Pirate Metrics](https://blog.agentanalytics.sh/blog/aarrr-for-solo-builders/): Use Agent Analytics with OpenClaw, Claude Code, Cursor, Codex, or any coding agent to diagnose which part of your growth loop is broken.
- [Why GPT-5.4 Makes Agent Analytics Better](https://blog.agentanalytics.sh/blog/gpt-5-4-openclaw-agent-analytics-growth-operator/): GPT-5.4 improved the capabilities that make AI-agent growth analysis work: long multi-step tasks, tool use, and polished business outputs.
- [Analytics Closes the Agent Feedback Loop](https://blog.agentanalytics.sh/blog/analytics-closes-the-agent-feedback-loop/): AI systems improve when action and consequence stay connected. Analytics is the measurement layer that tells your agent what happened after it shipped.
- [The Bullseye Method for Technical Indie Hackers](https://blog.agentanalytics.sh/blog/bullseye-method-for-technical-indie-hackers/): Use OpenClaw as an AI growth agent to explore channels, then use Agent Analytics to measure what actually drives activated users.
- [Advanced A/B Testing: Conditional Logic + Rich HTML Variants](https://blog.agentanalytics.sh/blog/advanced-ab-testing-conditional-rich-html/): Go beyond headline swaps. Test full sections, conditional experiences, and multi-step flows your AI agent can measure and improve.
- [Grow Your Projects from Claude Desktop](https://blog.agentanalytics.sh/blog/grow-your-projects-with-claude-desktop/): Use Claude Desktop as a growth agent to query analytics, spot bottlenecks, run experiments, and iterate without leaving your chat.
- [What's New in Tracker.js](https://blog.agentanalytics.sh/blog/whats-new-in-tracker-js/): Declarative events, opt-in browser signals, privacy guardrails, and a public browser smoke suite make the Agent Analytics tracker safer for Claude Code, Codex, Cursor, and your AI agent.
- [Stitch Users Across Sites](https://blog.agentanalytics.sh/blog/identity-stitching/): Identity stitching connects anonymous and signed-in behavior so your analytics, and the agents querying it, see one user journey.
- [Funnels: See Where Users Drop Off](https://blog.agentanalytics.sh/blog/funnels-where-users-drop-off/): Step-by-step conversion analysis your AI agent can query. Find the bottleneck, fix it, measure again.
- [A/B Testing Your AI Agent Can Actually Use](https://blog.agentanalytics.sh/blog/ab-testing-your-agent-can-use/): Run browser-side experiments with declarative HTML variants, no heavy SDK, and an API your agent can drive from setup to winner.
- [Talk to Your Analytics](https://blog.agentanalytics.sh/blog/talk-to-your-analytics/): Your AI agent can answer analytics questions from quick counts to multi-step analysis across experiments, funnels, and retention. No dashboards.
- [🦞 Set Up Agent Analytics with OpenClaw (5 Minutes)](https://blog.agentanalytics.sh/blog/setup-agent-analytics-with-openclaw/): Add analytics to all your projects and let your AI agent track what's working. No dashboards. Just ask.
- [Introducing the Agent Analytics Blog](https://blog.agentanalytics.sh/blog/introducing-agent-analytics-blog/): Updates, guides, and thoughts on analytics for products built, measured, and improved with AI agents.
- [How We Started Agent Analytics](https://blog.agentanalytics.sh/blog/how-we-started/): I was paying $28/mo to track pageviews across 5 side projects. I didn't want dashboards. I wanted my AI agent to handle it.