The smart way for AI Agents to access any URL. Cache-first structured JSON — no screenshots, no multimodal, 80%+ token savings. Fires automatically whenever your Agent needs to read a webpage.
Before (raw web_fetch)
"What's on pump.fun?"
→ fetch raw HTML → send to Claude
Cost: ~3000 tokens | Time: 8s
After (Smart Web Reader)
"What's on pump.fun?"
→ check EZCTO cache → structured JSON
Cost: 0 tokens | Time: 1s
A transparent web access layer that replaces raw HTML fetching with structured, cached intelligence.
Fires automatically whenever your Agent accesses any URL. No user action required — the Agent gets structured data instead of raw HTML.
Checks the EZCTO community library first — cache hits cost zero tokens. Local cache at ~/.ezcto/cache/ for instant repeat access.
Auto-detects crypto, ecommerce, restaurant, or general sites via pure text pattern matching. No LLM calls for detection.
Extracts image descriptions from alt text, context, filenames, and ARIA labels — no multimodal vision needed.
Every page you read is contributed back. The library grows with every user — popular sites become free for everyone.
JSON wrapped with metadata for skill chaining. Includes agent suggestions for next actions and skill combinations.
Fires automatically when your Agent accesses any URL. Cache hits cost zero tokens; misses are parsed locally and contributed back.
Agent needs to access a URL
Triggered automatically — no user action needed
Check EZCTO Community Library
GET api.ezcto.fun/v1/read?url=...
Return cached JSON
Instant · Zero tokens · Free
Agent reads structured data
Immediate understanding
Fetch HTML + detect site type
curl + zero-token pattern matching
Parse with local LLM
Structured JSON extraction
Validate + cache locally
~/.ezcto/cache/{hash}.json
Contribute to community library
Everyone benefits next time
Agent reads structured data
Ready for consumption
Agent needs to access a URL
Triggered automatically
Check EZCTO Community Library
GET api.ezcto.fun/v1/read?url=...
Explore the SKILL.md workflow instructions and the structured output format.
---
name: ezcto-smart-web-reader
description: Agent web access acceleration layer — reads any URL as structured JSON. Cache-first (public library hit = 0 tokens). The smart alternative to raw web_fetch.
---
# EZCTO Smart Web Reader
## What it does
Reads any URL and returns structured JSON containing page identity,
content sections, image text descriptions, video metadata, and
actionable links. Acts as the Agent's default web access layer —
replacing raw web_fetch with zero-token cache hits and intelligent
HTML parsing. 80%+ token savings vs screenshots.
## Triggers
Fires automatically whenever:
- Agent needs to read, access, or fetch a URL
- User provides a URL and wants to know what's on it
- Agent is about to use web_fetch to retrieve page content
- User asks to "look at", "check", "read", or "understand" a URL
## Workflow
1. Check EZCTO community library cache:
curl -s "https://api.ezcto.fun/v1/read?url={URL}"
→ If 200 + JSON → use directly (zero token cost)
2. Cache miss → fetch HTML:
curl -s -L -o /tmp/page.html "{URL}"
3. Compute HTML hash:
sha256sum /tmp/page.html
4. Auto-detect site type (pure text matching, zero tokens):
Scan HTML per references/site-type-detection.md
5. Assemble parsing prompt:
- Base: references/parse-prompt.md
- Extensions: references/extensions/{type}-fields.md
6. Parse with local LLM → output per references/output-schema.md
7. Validate JSON completeness
8. Dual-store:
- Local: ~/.ezcto/cache/{url_hash}.json
- Community library: POST /v1/contribute
9. Return structured JSON to Agent
## Guardrails
- Never modify URLs, addresses, or critical fields
- Truncate HTML > 500KB to <body> only
- Report errors explicitly, never guess content
- Never store or transmit API keysThree steps to give your Agent smart web access.
Visit clawhub.ai/takahashigy/ezcto-smart-web-reader and click Install. ClawHub handles everything automatically.
cd ~/.openclaw/skills
git clone https://github.com/pearl799/ezcto-web-translator ezcto-smart-web-readerAlternative: download the skill .zip above and extract to your OpenClaw skills directory.
openclaw skills list | grep ezcto-smart-web-readerThe SKILL.md file contains the workflow instructions. The references/ directory contains prompt templates and schemas.
Visit /api-keys to generate a free API key. This is only needed for server-side parsing — cache reads and contributions are always free.
# In OpenClaw chat — the skill fires automatically:
"What's on pump.fun?"
"Check this site: https://pancakeswap.finance"The skill intercepts URL access transparently. No special command needed.
Public endpoints at api.ezcto.fun — always-on, operated by the EZCTO team. Skill users do not need to deploy anything.
/v1/read?url={encoded_url}FREEQuery the community library for a cached result. Returns structured JSON if available. Always free, unlimited.
GET /v1/read?url=https%3A%2F%2Fpancakeswap.finance/v1/contributeFREESubmit a parsed result to the community library. Runs format + logic validation. Always free.
POST /v1/contribute
Content-Type: application/json
{
"url": "https://example.com",
"html_hash": "sha256_of_original_html",
"structured_data": { "meta": { ... }, ... },
"contributor_id": "user_abc123"
}/v1/read-serverAPI KeyServer-side parsing using EZCTO's LLM. Requires API key. Consumes 1 quota unit (3 if force_refresh).
POST /v1/read-server
Authorization: Bearer ezcto_xxxxxxxxxxxx
{
"url": "https://example.com",
"api_key": "ezcto_xxxxxxxxxxxx",
"force_refresh": false
}/v1/refreshAPI KeyForce re-parse a URL even if cached. Costs 3 quota units. Useful when website content has changed.
POST /v1/refresh
Authorization: Bearer ezcto_xxxxxxxxxxxx
{
"url": "https://example.com",
"api_key": "ezcto_xxxxxxxxxxxx"
}Cache reads and community contributions are always free. Pay only for server-side parsing when you need it.