Connect Claude Code, Claude Desktop or Codex to Shotomatic over MCP. Your agent can capture an ebook page by page, record a workflow, shoot a list of websites and export a PDF. You approve the agent, and Shotomatic asks again before anything is deleted.
Open the book at page one, capture every page of this window with the right arrow, and make a PDF.
Shotomatic brings the window forward, presses the key after each screenshot and stops by itself when the page stops changing.
Access is off until you connect an agent. After that, the agent works through the Shotomatic app on your computer.
Shotomatic lists the agents it finds on your computer: Claude (Claude Code and the Claude app) and Codex (ChatGPT).
Shotomatic adds itself to that agent's settings and turns access on. There is nothing to install and no command to type.
Start a new session and describe what you want captured. Claude Code and Codex work right away. Restart the Claude app to finish. None of them asks for approval again.
claude mcp add shotomatic -- npx -y shotomatic mcp{
"mcpServers": {
"shotomatic": {
"command": "npx",
"args": ["-y", "shotomatic", "mcp"]
}
}
}Add this to ~/Library/Application Support/Claude/claude_desktop_config.json, which you can open from Settings > Developer > Edit Config, then restart Claude Desktop.
Using another MCP client, or prefer to set it up by hand? These start the same bridge with npx, which needs Node 18.3 or later.
Describe the result you want. The agent picks the steps and uses the same Shotomatic features you do.
The agent captures one window, presses a key such as the right arrow after each screenshot, and stops when the page stops changing. Then it builds a Document and exports a PDF.
Capture every page of this Kindle window with the right arrow, then export the Document as a PDF.
The agent starts an Action Capture and you click through the task. Each click saves a step, and the agent can write a title and description for every step when you finish. Shotomatic does not click for the agent.
Record a guide for exporting a report from this app. I will click through it; write a title and description for each step when I say done.
Give the agent your pages and it captures them with your saved options or a Workflow, then reports which ones failed. Saving a Workflow with a schedule needs your confirmation, and scheduled Workflows run only while Shotomatic is open.
Capture these five pages on mobile, full page, and tell me which ones failed.
The agent can create a Document from a Capture Session, trim Pages, write descriptions and export a PDF, images or a ZIP. Files go to your export folder under new names, and existing files are never overwritten.
Trim the reader toolbar off every page of this Document, write a one-line description for each page, and export it as a PDF.
New to capturing book pages? Saving ebook pages as a PDF on a Mac
An agent gets only what you allow, and you can stop it at any time.
After you approve it, an agent can list your open windows and screens and capture the ones you point it to. It uses the same Screen Recording permission as Shotomatic, and a capture it starts shows the normal toolbar.
No. An agent has the same plan as you, so free-plan limits apply to it. Features that need Pro for you, such as some editing tools and longer stop conditions, need Pro for an agent too.
Shotomatic works with MCP clients such as Claude Code, Claude Desktop and Codex. We test with Claude Code and Codex. Another MCP client that can start npx -y shotomatic mcp should connect, but we have not checked other clients.
No. Shotomatic does not click for the agent. For an Action Capture you click through the task yourself, or use an agent that has its own computer-use ability.
If Shotomatic is not running when the agent connects, the bridge opens it in the background. Scheduled Workflows run only while the app is open.
Shotomatic keeps them on your computer and opens no network port. If an agent looks at an image, that image becomes part of the agent's conversation and is handled under the agent's own terms.
Download Shotomatic, open Settings > AI agents, and connect the agent you already use.