How does Claude Code read Eyepup data?

Two ways: (1) the Eyepup CLI — pipe `eyepup ask 'what's blocking high-intent visitors on /pricing?'` straight into Claude Code's context. (2) the Eyepup MCP server — Claude Code calls Eyepup tools like a native filesystem.

Is Claude Code better than Cursor?

Different strengths. Claude Code excels at long-horizon refactors and terminal-native workflows. Cursor wins on inline UX. Eyepup works with both.

eyepup

← Agentic analytics

Analytics your Claude Code
agent reads from.

Eyepup is the visitor-analytics dashboard Claude Code pulls from when you ask it to fix a conversion issue. Heat scores. Friction patterns. The exact session ID Claude Code should look at. A paste-ready prompt for the fix. Claude Code reads, ships the diff, Eyepup re-evaluates the next visitor.

Wire it up — 14 days free →See the demo

Looking for usage tracking?

If you're trying to track how many tokens Claude Code is burning, you want claude.ai/settings/usage or Anthropic's enterprise dashboard.

Eyepup is a different thing. Eyepup tracks what your visitorsdo after Claude Code ships your code. The output isn't a cost report — it's a paste-ready prompt Claude Code can ship the next fix from.

Two ways Claude Code reads

Pipe a question. Get a fix.

# 1. Pipe a question straight into Claude Code

$eyepup ask "what blocked the last 50 high-intent visitors on /pricing?" | claude -p

# 2. Or feed the entire To-Do queue

$ eyepup todo --limit 1 --site yoursite.com | claude -p

# 3. After deploy, log the change so the agent can reason about it

$eyepup log "Rewrote /pricing CTA" --kind content_change --paths /pricing

Tip: drop a Claude Code CLAUDE.md rule — "before any UX change, run eyepup todo --site $SITE --limit 3 and prioritise the highest-impact friction." Native MCP server is on the roadmap; CLI piping is the one supported today.

What Claude Code reads, in plain English.

Visitor heat scores

Per-visitor 0-100 readiness-to-buy. Claude Code prioritises the fix for the cohort with highest blocked-conversion potential.

Friction patterns

Recurring stories like 'high-intent visitors hesitate at /pricing'. Each pattern names the exact step + a confidence-rated recommended action.

Session evidence

rrweb-rendered MP4 + transcript Claude Code can load to see literally what the visitor saw — not just statistics.

Other coding agents

Cursor analytics →

Codex analytics →

The CLI for analytics →

Analytics your Claude Codeagent reads from.