MCP Server

Rendex provides a Model Context Protocol server so AI agents can render HTML and capture screenshots natively.

Remote Server (Recommended)

Connect directly to mcp.rendex.dev — no installation needed.

{
  "mcpServers": {
    "rendex": {
      "url": "https://mcp.rendex.dev/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

ChatGPT, Claude, and other OAuth-capable clients can connect with no API key — the remote server is fronted by OAuth 2.1 (authorization code + PKCE). You sign in once with your Rendex email and each agent renders against your own plan and credit pool. Clients that prefer a static key can pass Authorization: Bearer rdx_… as shown above.

Local via npx

Run the MCP server locally using stdio transport:

{
  "mcpServers": {
    "rendex": {
      "command": "npx",
      "args": ["-y", "@copperline/rendex-mcp"],
      "env": {
        "RENDEX_API_KEY": "YOUR_API_KEY"
      }
    }
  }
}

Claude Code (CLI)

{
  "mcpServers": {
    "rendex": {
      "command": "npx",
      "args": ["-y", "@copperline/rendex-mcp"],
      "env": {
        "RENDEX_API_KEY": "YOUR_API_KEY"
      }
    }
  }
}

Add to ChatGPT or Claude

Prefer the chat app? Add Rendex as a custom connectorin ChatGPT or Claude and authenticate once with OAuth — no API key to paste. Each agent then renders against your own plan and credits.

https://mcp.rendex.dev/mcp

Available Tools

The Rendex MCP server registers 13 tools— five rendering tools and the eight Rendex Watch change-monitoring tools. Each is annotated read-only or destructive/idempotent so a client can reason about safety.

rendex_screenshot

Capture a screenshot or PDF of any webpage or raw HTML.

Source

Output

Viewport

Timing

Content Blocking

CSS & JS Injection

Cookies & Headers

PDF Options

These parameters only apply when format="pdf".

Geo-Targeting

Render the page as seen from a specific country. Requires a Pro or Enterprise plan.

Async Processing

rendex_extract

Extract clean reader-mode content from any webpage as Markdown, JSON, or HTML. Runs the same Chromium render pass as a screenshot, so it captures content after JavaScript runs \u2014 handles SPAs that fetch-only readers miss. Strips nav, ads, and boilerplate, returning the article body plus title, byline, and excerpt. Great for feeding page content to an LLM, summarization, or RAG ingestion. Available on every plan, including Free.

Returns the extracted content as a text block. For image plus text in one call, use rendex_screenshot on POST /v1/screenshot/json with extract: true.

render_artifact

Turn Markdown or HTML into a branded, downloadable artifact — a PDF, a PNG, and a hosted share page — in one call. Apply a logo, accent color, font, header, and footer, and choose the PDF page size, orientation, and margins. Returns hosted URLs { pdfUrl, pngUrl, shareUrl, expiresAt }. Each requested format costs 1 render credit.

rendex_render_link

Render a URL, HTML, or Markdown and get back a signed, hosted, edge-cached image URL instead of the bytes — ideal for dynamic og:image tags. Takes the same parameters as rendex_screenshot, plus an optional expiresIn (60–2592000s). Returns { url, expiresAt, format, cacheTtl }. Costs 1 render credit per fresh render; cached repeat hits don't re-charge.

rendex_account

Check the account from inside the agent: the current plan, this month's usage (used / limit / remaining and the reset date), the per-minute rate limit, and a one-tap upgrade link. Takes no parameters. Read-only — costs no credits.

Rendex Watch tools

Rendex Watchmonitors a page for changes on a schedule; over MCP the whole lifecycle is available as eight tools, so an agent can stand up, inspect, retune, and tear down a monitor in conversation — no dashboard. Handlers return watch metadata and signed before/after/overlay image URLs as JSON text. Plan walls (interval floor, watch cap, webhooks = Starter+, geo & AI summary = Pro+) surface as tool errors. See the Watch docs for the full change-detection model.

MCP workflow recipes

Step-by-step agent walkthroughs of these tools — screenshotting, extracting, watching, and rendering share links — each source-accurate to a real MCP tool. Browse the full set on the recipes hub.

Response Format

rendex_screenshot returns two content items:

1. Image: Base64-encoded screenshot as an image content block
2. Metadata: JSON text block with URL, dimensions, format, and file size

rendex_extract returns the cleaned content (title, site name, and body) as a single text block.

Example Usage in Claude

Once connected, simply ask:

"Take a screenshot of https://example.com in dark mode"

Claude will call the rendex_screenshot tool and display the captured image directly in the conversation.

Not Using AI Agents?

If you're building a regular application (not an AI agent), use the official SDKs instead of MCP:

See the Quick Start for SDK examples.

Related

Building an agent that captures or reads web pages? See screenshots for AI agents for the capability deep-dive, or browse workflow recipes for end-to-end blueprints.