AI-native

Built for AI agents. Not retrofitted.

Simple API that any AI agent can call. Structured JSON responses, predictable errors, and tool-use ready endpoints.

Agent output

What your agent sees

Your agent calls the API, gets back structured JSON with the screenshot. These were captured in a single batch call.

Screenshot of stripe.com
stripe.com650ms
Screenshot of github.com
github.com430ms
Screenshot of vercel.com
vercel.com560ms

Why agents love it

Designed for machine consumption

Structured responses

Every response is typed JSON with metadata — dimensions, file size, render time, cache status. No binary parsing, no guessing.

Predictable errors

Standard HTTP status codes with structured error bodies. Agents handle failures programmatically without parsing error strings.

Tool-use ready

Designed as an OpenAI/Anthropic tool definition. One function call, one screenshot. Works with any agent framework.

Tool definition

Maps directly to a tool call

The API is designed to work as a tool definition for any AI agent framework. One function, clear parameters, structured response.

Copy this definition into your agent's tool list and it just works — Claude, GPT-4, LangChain, or any custom agent.

tool-definition.json
{
  "name": "take_screenshot",
  "description": "Capture a screenshot of any URL",
  "parameters": {
    "type": "object",
    "properties": {
      "url": {
        "type": "string",
        "description": "The URL to capture"
      },
      "format": {
        "type": "string",
        "enum": ["png", "jpeg", "webp"],
        "default": "webp"
      },
      "full_page": {
        "type": "boolean",
        "default": false
      }
    },
    "required": ["url"]
  }
}

Compatibility

Works with every agent framework

If your agent can make HTTP requests, it can take screenshots.

OpenAI function calling

Works as a tool with GPT-4, GPT-4o, and any OpenAI-compatible model.

Anthropic tool use

Native tool_use format for Claude. MCP server available.

LangChain

Drop-in tool for LangChain agents. Python and TypeScript.

CrewAI

Custom tool integration for multi-agent workflows.

AutoGPT / AutoGen

HTTP tool that any autonomous agent framework can call.

Custom agents

Simple REST API. If your agent can make HTTP requests, it works.

Batch processing

Batch + async for pipelines

Queue up to 100 URLs in a single API call. Get results via webhook as each screenshot completes. Built for data pipelines and web research agents.

POST /v1/batch
{
  "urls": [
    "https://example.com",
    "https://stripe.com",
    "https://github.com"
  ],
  "format": "webp",
  "webhook_url": "https://your-app.com/webhook"
}

Give your agents eyes

One API call. Structured response. Any agent framework.