Screenshot API for ChatGPT Custom Actions

Add visual web perception to any Custom GPT or OpenAI Assistant — copy the schema below and paste it into the Actions editor.

How It Works

Custom GPTs and OpenAI Assistants support Actions — custom API calls defined by an OpenAPI schema. When a user asks your GPT to screenshot a URL, it calls the Hermesforge API and returns the image directly in the conversation.

ChatGPT already routes screenshot requests to this API organically — many users have discovered it without any integration work. Adding it as an explicit Action gives you control over the behavior and unlocks structured use cases.

Quick Setup (2 minutes)

https://hermesforge.dev/openapi/screenshot.json

Or use the full multi-tool schema (screenshot + tech stack + dead link checker + SEO audit) at https://hermesforge.dev/openapi.json.

Manual Schema

If you prefer to paste the schema directly, use this minimal version covering the screenshot action:

Copy{
  "openapi": "3.1.0",
  "info": {
    "title": "Hermesforge Screenshot API",
    "description": "Capture any URL as PNG, JPEG, WebP, or PDF. Supports dark mode, full-page, mobile viewport, ad blocking, and custom JavaScript.",
    "version": "2.1.0"
  },
  "servers": [
    { "url": "https://hermesforge.dev" }
  ],
  "paths": {
    "/api/screenshot": {
      "get": {
        "operationId": "captureScreenshot",
        "summary": "Capture a screenshot of any URL",
        "description": "Renders the page in a real Chromium browser and returns an image. Free tier: 50/day, no API key needed. Optional: pass X-API-Key header for higher limits.",
        "parameters": [
          {
            "name": "url",
            "in": "query",
            "required": true,
            "description": "The URL to screenshot (must include https://)",
            "schema": { "type": "string", "example": "https://example.com" }
          },
          {
            "name": "format",
            "in": "query",
            "required": false,
            "description": "Output format: png (default), jpeg, webp, or pdf",
            "schema": { "type": "string", "enum": ["png", "jpeg", "webp", "pdf"], "default": "png" }
          },
          {
            "name": "viewport",
            "in": "query",
            "required": false,
            "description": "Device preset: mobile, tablet, desktop (default: desktop 1280x720)",
            "schema": { "type": "string", "enum": ["mobile", "tablet", "desktop", "desktop_hd", "macbook", "iphone_14", "og", "twitter"] }
          },
          {
            "name": "full_page",
            "in": "query",
            "required": false,
            "description": "Capture full scrollable page (true) or viewport only (false)",
            "schema": { "type": "boolean", "default": true }
          },
          {
            "name": "dark_mode",
            "in": "query",
            "required": false,
            "description": "Enable dark mode rendering",
            "schema": { "type": "boolean", "default": false }
          },
          {
            "name": "block_ads",
            "in": "query",
            "required": false,
            "description": "Block ads, trackers, and cookie banners",
            "schema": { "type": "boolean", "default": false }
          },
          {
            "name": "width",
            "in": "query",
            "required": false,
            "description": "Viewport width in pixels (ignored if viewport preset set)",
            "schema": { "type": "integer", "default": 1280 }
          },
          {
            "name": "delay",
            "in": "query",
            "required": false,
            "description": "Wait N milliseconds after page load before capturing (useful for animations or lazy-loaded content)",
            "schema": { "type": "integer", "default": 0 }
          }
        ],
        "responses": {
          "200": {
            "description": "Screenshot image (PNG, JPEG, WebP) or PDF",
            "content": {
              "image/png": { "schema": { "type": "string", "format": "binary" } },
              "image/jpeg": { "schema": { "type": "string", "format": "binary" } },
              "image/webp": { "schema": { "type": "string", "format": "binary" } },
              "application/pdf": { "schema": { "type": "string", "format": "binary" } }
            }
          },
          "429": {
            "description": "Rate limit exceeded. Get a free API key at https://hermesforge.dev/api/keys"
          }
        }
      }
    }
  }
}

Example GPT Prompt Instructions

Add these instructions to your Custom GPT's system prompt to enable natural-language screenshot requests:

CopyYou can capture screenshots of any URL using the captureScreenshot action.

When a user asks to "screenshot", "capture", "show me", or "take a picture of" a URL, call captureScreenshot with:
- url: the full URL (always include https://)
- viewport: "mobile" for mobile view, "desktop" for desktop (default)
- full_page: true unless the user specifies viewport/above-fold only
- dark_mode: true if user asks for dark mode or night mode
- block_ads: true if user asks for clean or ad-free screenshot

Display the returned image inline in your response.

If you receive a 429 error, tell the user: "The free tier limit has been reached. They can get a free API key at https://hermesforge.dev/api/keys to continue."

Use Cases

Rate Limits & API Keys

API keys are free for the Starter tier. Get one at /api/keys. Pass as X-API-Key: hf_yourkey header or &key=hf_yourkey query parameter.

For production Custom GPTs that need consistent availability, add your API key to the Action's authentication settings (type: API Key, header: X-API-Key).

Already Working?

ChatGPT's ChatGPT-User bot already routes screenshot requests to this API through organic discovery. If your Custom GPT is generating those calls, you're already integrated — this page just makes it explicit and gives you control over the parameters.

Other Integrations

LangChain Tool — Full OpenAPI Spec — Plugin Manifest — API Reference