Skip to content

Connect a coding agent

Coding agents such as Claude Code, Cursor and VS Code can read your library the same way Claude Desktop does. ShotWork has no one-window route for them, so you paste one entry into the agent's own settings instead.

Before you start

  • An agent that reads its connections from a settings file. Its own documentation says where that file is and what it calls the section the entry goes in.
  • A license. See The free tier and your license.
  • The shot selects you want it to see, listed in the Library window.

Steps

  1. Choose File › Connect an AI Assistant….
  2. On the Claude Desktop tab, choose Set it up by hand ▸ at the bottom of the window. The box that opens belongs to the tab you are on, and Claude Desktop's is in the shape most agents read.
  3. Copy the entry named shotwork from the box.
  4. Open your agent's settings file and add the entry inside its mcpServers section. If the section is not there, create it. An agent that calls the section something else wants the same entry under its own name.
  5. Restart the agent.
  6. Ask it: What's in my ShotWork library?

The box opens with Claude Desktop's settings path, because that is the file ShotWork would write for you. For a coding agent, ignore the path and take the entry.

What happens

The agent starts ShotWork's library server itself whenever it needs your library, and asks it the same things Claude Desktop does. ShotWork does not need to be open, and the same checks apply: restricted shots stay out unless you ask, and everything built lands in ShotWork Reels.

What the entry contains

The agents speak MCP, the Model Context Protocol, the standard the assistant apps use to talk to local programs. The entry tells the agent how to start ShotWork's server:

{
  "mcpServers": {
    "shotwork": {
      "command": "C:\\Users\\artist\\AppData\\Local\\Programs\\ShotWork\\ShotWork.exe",
      "args": ["--mcp-server"]
    }
  }
}

command is your installed ShotWork; the box in the window fills in the real path. When the server starts, it writes three lines to the agent's log, profile:, library: and reels:, naming the profile it runs on, the library it reads and the folder it writes to.