# AI Agent Integrations — Give Your AI the Power to Browse

> Markdown version of https://capture.page/ai-agents
> Part of Capture, the web automation platform for humans and agents.
> Pricing: https://capture.page/pricing.md

Three ways to connect Capture to your AI workflows. Screenshots, PDFs, content
extraction, and metadata — all accessible through the tools your agents already
use.

## 1. MCP Server (Model Context Protocol)

Connect Capture directly to AI tools like Claude Code, Claude Desktop, Cursor,
and Codex. Generate screenshots, PDFs, and extract content without leaving your
conversation.

- Server URL: `https://capture.page/mcp/v1` (Streamable HTTP transport)
- Authentication: `Authorization: Bearer YOUR_TOKEN`
- Docs: https://docs.capture.page/docs/mcp-integration

Tools:
- `capture_screenshot` — full-page screenshots with device emulation, dark mode, ad blocking, and custom viewports
- `capture_pdf` — generate PDFs with custom page sizes, margins, orientation, and background graphics
- `capture_content` — extract HTML and cleaned text content from any website
- `capture_metadata` — extract title, description, Open Graph tags, and publisher info

Quick add (Claude Code):

```bash
claude mcp add \
  --transport http capture \
  https://capture.page/mcp/v1 \
  --header "Authorization: Bearer YOUR_TOKEN"
```

## 2. AI Agent Skill

Install Capture as a skill for any agent that supports skills. Ask in natural
language and the skill handles the rest.

- Docs: https://docs.capture.page/docs/ai-agent-skill
- Natural language — e.g. "take a screenshot of example.com"
- Device emulation, content extraction (markdown/HTML), and animated GIF recordings

```bash
npx skills add techulus/capture-skills

# Then just ask your agent:
# "Take a screenshot of example.com"
# "Generate a PDF of this article"
# "Extract content from this page as markdown"
```

## 3. Capture CLI

A CLI built in Go that any AI agent or pipeline can invoke. Pipe web captures
into automation pipelines, CI/CD workflows, or agent tool chains.

- Docs: https://docs.capture.page/docs/capture-cli
- Use `--edge` for faster responses from the nearest edge location

```bash
# Install via Homebrew
brew tap techulus/tap && brew install capture

# Capture a screenshot
capture screenshot https://example.com -o shot.png

# Generate a PDF
capture pdf https://example.com -X format=A4 -o doc.pdf

# Extract content as markdown
capture content https://example.com --format markdown
```

## How it works

1. **Connect** — add your Capture credentials (an API key for the CLI and Skill, or a Bearer token for MCP). Takes under a minute.
2. **Ask** — use natural language or direct commands: "screenshot this page", "extract the content as markdown", or pipe it into a script.
3. **Get results** — Capture renders in real time on a global edge network; screenshots, PDFs, content, and metadata delivered in seconds.

## Use cases

- **Research agents** — browse the web, extract article content, and gather structured data from any page without a headless browser.
- **Documentation pipelines** — screenshot UI states, generate PDF reports, and capture page metadata as part of a docs workflow.
- **QA & monitoring** — schedule visual regression checks, capture before/after screenshots, verify page content programmatically.
- **Content generation** — feed real-time web content to LLMs; extract clean markdown for summarization, analysis, or RAG pipelines.
- **Sales & outreach** — capture personalized screenshots of prospect sites, generate PDF proposals, enrich CRM data with metadata.
- **CI/CD integration** — add visual snapshots to deployment pipelines; verify rendered output, catch regressions, archive page states.

## See also

- Browser Sessions (stateful browsing for agents): https://capture.page/browser-sessions.md
- API overview for agents: https://capture.page/llms.txt
- Pricing: https://capture.page/pricing.md