Skip to content

MCP Tools

Claude Yard exposes its functionality as Model Context Protocol (MCP) tools. This lets any MCP-compatible client — including Claude Code itself — spawn sessions, send keystrokes, open browsers, and manage the layout.

ToolWhat it does
spawnCreate a new terminal session (shell or Claude)
send_keysSend keystrokes to a terminal panel
get_stateGet current panels, sessions, and layout
set_layoutArrange panels in the workspace
open_browserOpen a browser panel with a URL

Additional tools may be available depending on which modules are loaded. Run node cli.mjs tools to see the full list.

The MCP server uses stdio transport. Configuration depends on your client.

Add to your Claude Code MCP config (~/.claude/claude_code_config.json or project-level .mcp.json):

{
"mcpServers": {
"claude-yard": {
"command": "node",
"args": ["/path/to/volter/mcp-server.mjs"]
}
}
}

Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):

{
"mcpServers": {
"claude-yard": {
"command": "node",
"args": ["/path/to/volter/mcp-server.mjs"]
}
}
}

Any client that supports stdio transport can connect. The server binary is mcp-server.mjs in the Claude Yard installation directory. Point your client’s MCP config at it with node as the command.

MCP Clientstdiomcp-server.mjsproxyHTTPClaude Yardtools/list → GET /api/mcp/tools    tools/call → POST /api/tools/call

The MCP server is a thin proxy — it translates MCP protocol messages into HTTP calls to the local Claude Yard server. This means it always reflects the current set of available tools, including any added by custom modules.

With MCP, you can set up workflows where one Claude instance manages others:

  • Parallel task execution — spawn multiple sessions and distribute work across them
  • Automated testing — open a browser, run tests in a terminal, and verify results
  • Pipeline orchestration — chain sessions together where one Claude’s output feeds another’s input

See Workflows for concrete examples.