Agent Analytics AgentAnalytics Agent-ready analytics
Works with
Claude Code Cursor Paperclip Codex OpenClaw

Web analytics your
AI agent can run.

Create projects, add tracking, query stats, monitor sites, and run experiments through plugin, skill, MCP, CLI, or API instead of living in dashboards.

Start free Choose your agent
100k free events/month · 2 projects $1 per 10k events on Cloud Pro
Plugin
Skill
MCP
CLI / npx
API
Outcome loop
01
Step 01

Work ships

A landing page changes. A docs flow gets updated. A launch post goes live.

02
Step 02

Results measured

Agent Analytics turns shipped work into outcome signals your AI agent can read and compare.

Traffic +22% Signup flat Activation +14% Retention unchanged
03
Step 03

Next move

Keep the onboarding winner. Rework the post angle. Investigate the retention leak.

Any AI agent can ship work. The advantage comes from measuring outcomes and using them in the next decision.
What this is

One analytics surface, exposed everywhere your agent already works.

Agent Analytics is web analytics built for Claude Code, Paperclip, Codex, OpenClaw, Cursor, Claude Desktop, and similar tools. Instead of forcing a dashboard workflow, it exposes the same analytics surface through plugin, skill, MCP, CLI, and raw API access.

Keep the browser for billing. Run the day-to-day operating loop through code, terminals, and chat.

How Agent Analytics is shaped for agents:

01

Projects are API objects

Projects, tokens, stats, and experiments are available programmatically instead of being buried inside a workspace UI.

02

The same surface works across access paths

Use a plugin in Claude Code, a skill in Codex, MCP in chat tools, or raw HTTP in custom runtimes without changing the underlying workflow.

03

Workflow layers meet agents where they live

Plugins and skills package the common setup, monitoring, and experiment loops so your agent can stay inside its native environment.

The core idea: your agent should not need a dashboard-only mode just to operate web analytics.

Use cases

The jobs your AI agent can own once web analytics becomes part of its toolchain.

Traditional analytics assumes a human will click through reports. Agent Analytics gives your AI agent a machine-readable surface so it can set up projects, watch them, and act on what it finds.

Ship and verify

Launch analytics without leaving the build loop

Tell your agent to set up analytics for a new site and it can create the project, return the snippet, and confirm the first event before the task is done.

  • The prompt starts with the app you just shipped, not a dashboard walkthrough.
  • The agent chooses the right install path for the environment it already uses.
  • You get proof that tracking is live before moving on.
Morning brief

Check every project in one pass

Ask your agent what changed this morning and it can scan all your projects, flag anomalies, and summarize what actually needs attention.

  • The prompt is a plain-language check-in, not a reporting ritual.
  • The agent compares projects, reads spikes, and filters out bot-heavy noise.
  • You get one short brief instead of tab-hopping.
Find conversion leaks

Turn suspicious numbers into concrete fixes

When traffic or conversion shifts, your agent can inspect funnels, pages, sessions, sources, and bot traffic to isolate where users drop off.

  • The prompt starts from the business symptom, not the chart type.
  • The agent follows the leak across pages, sources, and devices.
  • You get the likely broken step and what to inspect next.
Run growth experiments

Operate experiments from the same agent surface

Your agent can create, QA-check, read, and close CTA or landing-page experiments without sending you into a separate product.

  • The prompt starts with the hypothesis you want to test.
  • The agent wires the experiment, verifies variants, and reads the result.
  • You get a recommendation on whether to keep running, stop, or ship the winner.
Once your agent is connected

These are machine-usable workflows, not dashboard rituals your team has to remember to perform.

Step 01

Ship a new site with analytics already wired

Tell your agent "set up analytics for this app" and it can create the project, return the snippet, and confirm the first event.

Step 02

Read the morning brief

Ask what changed across all your projects and your AI agent can surface anomalies, traffic swings, and the one thing worth attention now.

Step 03

Inspect funnels and experiments

Ask where users drop off or which variant is winning. Agent Analytics returns the analytics surface while your agent interprets it.

Step 04

Pick the next move

When enough data exists, your AI agent can summarize the result and point you to the next page, funnel step, or experiment worth touching.

Choose your agent

One analytics surface. Native entrypoints for each environment.

Start with the path your agent already supports. The underlying analytics surface stays the same; only the native entrypoint changes.

Plugin-first

Claude Code

Use the plugin when you want the MCP connection and the analytics workflow layer in one install.

  • Shortest hosted path inside Claude Code.
  • Bundles MCP connectivity with Agent Analytics workflow guidance.
Hosted MCP

Claude Desktop / Cowork

Connect the hosted MCP server when you want native tool calls inside a chat-first environment.

  • Best default for connector-style chat tools.
  • Keeps auth and tool access inside the hosted flow.
Skill + plugin

Paperclip

Delegate setup to your CEO in Paperclip, give that chain the Agent Analytics skill, and let it run. In the normal path, one browser approval is enough for the agent to finish login, create the project, and verify the first event.

  • Fast setup: one CEO task, one browser approval, and the first real analytics read from the same issue flow.
  • Add the live plugin only when you also want a shared company-level live map inside the workspace.
Setup guide Why Paperclip needs this
Skill + CLI

Cursor

Start with the Agent Analytics skill plus CLI workflows, and add MCP only when you specifically want connector-style tool calls.

  • Lower overhead for shell-capable editor agents.
  • Best fit when the agent can already execute commands.
Skill-first

OpenAI Codex

Install the skill when you want Codex to query projects, inspect health, and run experiments from the same conversation loop.

  • Keeps the workflow agent-native without requiring MCP.
  • Best path when the environment already supports `npx` and skills.
Skill-first

OpenClaw

Install the skill through ClawHub and let OpenClaw run project setup, reporting, and experiment workflows from chat.

  • Cleanest path for scheduled analytics jobs from chat.
  • Best fit when OpenClaw already owns the command loop.
CLI + MCP + API

Any AI agent

If your AI agent supports skills, CLI commands, MCP, or direct API calls, it can work with Agent Analytics.

  • Use the native entrypoint your environment already supports.
  • Same analytics surface whether you come through a skill, CLI, MCP, or raw API.
Choose your setup

Choose deployment first, then use the access path that fits your agent. Cloud gives the fullest hosted agent surface; self-host keeps web analytics on your infrastructure.

Open Source

Self-hosted

Best for teams that want API- and CLI-driven web analytics on their own infrastructure.

Price
Free forever
Hosting
Your infra
Projects
Self-managed
Events
No hosted metering
Agent/API reads
No hosted metering
Retention
You control retention
Agent access
API, CLI, and direct integration on your infra. No hosted connector layer.
Tracker
Basic tracker included. Advanced tracker available when enabled in tracker.js.
Analytics access
Core analytics surface: project list, stats, recent events
Experiments
Not included
Self-host open source
Cloud Free

Start free

Hosted agent analytics for testing the zero-dashboard workflow on your first couple of projects.

Price
$0
Hosting
Hosted by Agent Analytics
Projects
2
Events
100k / month
Agent/API reads
500 / month
Retention
90 days
Agent access
Plugin, skill, CLI, and API are available. Hosted MCP is constrained by free-tier limits.
Tracker
Basic tracker included. Advanced tracker not included.
Analytics access
Core analytics surface: project list, stats, recent events
Experiments
Not included
Start free
Cloud Pro

Pay as you go

Full hosted agent operating surface for teams that want unlimited projects, deeper signals, and experiments.

Price
$1 per 10k events
Hosting
Hosted by Agent Analytics
Projects
Unlimited
Events
No monthly cap
Agent/API reads
No monthly cap
Retention
365 days
Agent access
Full hosted agent surface: plugin, skill, MCP, CLI, and API.
Tracker
Full tracker signals: impressions, scroll depth, clicks, forms, errors, performance, web vitals
Analytics access
Full analytics surface: funnels, retention, sessions, pages, heatmaps, insights, live
Experiments
Included
Start hosted

All plans are agent-usable. The best access path depends on your environment and deployment model.

Open source or hosted

Give your agent web analytics it can actually operate.

Start free on our cloud or self-host the open-source stack. Either way, the day-to-day loop stays in code, chat, and terminals instead of dashboard tabs.

Start free Self-host open source