Clip the Web Into
Clean Markdown for Your AI
Zero token tax. Zero server transit. One-click capture to Obsidian, Feishu/Lark, your own webhook, or a local MCP (Model Context Protocol) server.
View Video Transcript
This 15-second demo shows the Developer Gateway workflow:
- Open the EasyCatch sidebar on any webpage you want to capture.
- Choose Markdown mode to extract clean, structured text with ads and navigation removed.
- Export the clip to Obsidian, a Lark/Feishu Bitable, or a custom webhook endpoint.
- Optionally trigger captures from Claude Code or Cursor through the local MCP server.
Built for AI Developers
Researchers, knowledge gardeners, and MCP hackers — this is your lane.
📝 Clean Markdown Pipeline
Turn any webpage into structured, AI-ready Markdown with the noise stripped out. Perfect for context windows, RAG, and knowledge bases.
- Semantic Extraction — headings, tables, lists, and code blocks preserved, ads and nav removed.
- Zero Token Tax — only the content that matters reaches your LLM context window.
- Local-First — all processing happens in your browser; nothing leaves your machine.
- Obsidian Ready — one-click export to your vault with frontmatter and canonical links.
🔌 Automation & Integrations
Wire EasyCatch into your toolchain. Webhooks, team spreadsheets, or an MCP server — you choose the destination.
- Webhook Ready — POST clean JSON to n8n, Make, Zapier, or any custom endpoint.
- Lark / Feishu Bitable — auto-append captures to team tables for research and review.
- Local MCP (Model Context Protocol) Server — expose EasyCatch as an MCP tool to Claude Code, Cursor, or any MCP client.
- Open JSON Schema — predictable payload shapes for parsing and downstream transformation.
Plug Into Your Stack
Three lines of config, endless automation possibilities. Pick your integration.
Webhook Endpoint
Captured data is POSTed as JSON to your endpoint. Works with n8n, Make, Zapier, or any HTTP server.
POST /your-webhook-endpoint
Content-Type: application/json
{
"title": "Product Name",
"price": "$12.99",
"images": ["url1", "url2"],
"specs": { "Material": "Cotton", "Size": "XL" },
"source_url": "https://detail.1688.com/...",
"captured_at": "2026-06-01T12:00:00Z",
"format": "markdown"
}Webhook & MCP Output Structure
Predictable payloads for automation tools, knowledge bases, and AI clients.
| Output Channel | Format | Key Fields / Behavior |
|---|---|---|
| Webhook | JSON POST | title, url, markdown, images[], specs[], capturedAt, plus MCP Sentinel signature header. |
| Lark / Feishu Bitable | Row append | title, url, summary, tags; appended locally from your browser using your Personal Access Token. |
| Obsidian | Markdown file | YAML frontmatter with title, source, capturedAt, plus cleaned body and canonical link. |
| MCP Tool Call | JSON-RPC via local MCP server | tool name, arguments object, and MCP Sentinel verification to authenticate the client. |
Model Context Protocol (MCP) vs Traditional Webhooks
Choosing between a local-first MCP server and a cloud webhook stream for automated Shopify sourcing, DOM sniffing, and CSV feeds.
| Dimension | Local-First MCP | Traditional Webhook |
|---|---|---|
| Round-trip latency | Localhost — no network hop | Cloud delivery — network round-trip |
| Anti-bot / proxy evasion | None needed — runs in your browser | Often required for source-site scraping |
| Data ownership | Captured data stays on your device | Payload transits a third-party server |
| Integration surface | JSON-RPC tool calls to your local agent | HTTP POST to your endpoint |
| Fail-open / offline behavior | Works offline, deterministic capture | Depends on external endpoint availability |
| Best fit | Local AI agents, Obsidian, MCP clients | Server-side pipelines, Lark/Feishu Bitable |
Because EasyCatch's MCP server runs locally on your machine, it is the low-latency path for e-commerce sourcing: a capture round-trip between your browser and the local JSON-RPC server avoids the cloud entirely, cutting typical delivery latency by up to 95% compared with webhook calls that hop to a remote endpoint and back. That same local architecture also eliminates proxy rotation and TLS fingerprint evasion — the source-site anti-bot countermeasures that cloud scraping services must build, maintain, and pay for. A cloud webhook remains the right tool when you want a server-side pipeline (for example, appending rows to Feishu/Lark Bitable), but for AI-driven selection workflows the local MCP server is the faster and more resilient option.
Developers can pipe every capture through the MCP server config or a webhook event stream — both receive the same clean Markdown, images, and spec fields.
Developer FAQ
Common questions about integrations, security, and configuration.
What is MCP and how does EasyCatch use it?
MCP (Model Context Protocol) is an open standard that lets compatible AI clients call local servers as tools. EasyCatch exposes a local MCP server so Claude Code, Cursor, and other MCP clients can trigger captures. Data never leaves your machine; requests are verified with the MCP Sentinel protocol.
Is captured data secure?
Yes. EasyCatch is local-first: extraction, encryption, and storage happen inside your browser. No backend servers process your captured content. Pro license verification uses a minimal authentication server only to validate your License Key. We do not track or store your content.
Can I send clips to Obsidian?
Yes. Pro users can export clean Markdown with frontmatter and canonical links directly into an Obsidian vault.
Does the Lark/Feishu integration require a server?
No server is required. You provide your own Lark/Feishu Personal Access Token and Bitable IDs; the extension appends rows locally from your browser.
What format does the webhook payload use?
Webhook payloads are JSON and include title, price, images, specs, source URL, capture timestamp, and format. They work with n8n, Make, Zapier, or any HTTP endpoint.