registry/registry/coder-labs/modules/cursor-cli/README.md

display_name	icon	description	verified	tags
Cursor CLI	../../../../.icons/cursor.svg	Run Cursor Agent CLI in your workspace for AI pair programming	true	agent cursor ai tasks

Cursor CLI

Run the Cursor Agent CLI in your workspace for interactive coding assistance and automated task execution.

module "cursor_cli" {
  source   = "registry.coder.com/coder-labs/cursor-cli/coder"
  version  = "0.2.1"
  agent_id = coder_agent.main.id
  folder   = "/home/coder/project"
}

Basic setup

A full example with MCP, rules, and pre/post install scripts:

data "coder_parameter" "ai_prompt" {
  type        = "string"
  name        = "AI Prompt"
  default     = ""
  description = "Build a Minesweeper in Python."
  mutable     = true
}

module "coder-login" {
  count    = data.coder_workspace.me.start_count
  source   = "registry.coder.com/coder/coder-login/coder"
  version  = "1.0.31"
  agent_id = coder_agent.main.id
}

module "cursor_cli" {
  source   = "registry.coder.com/coder-labs/cursor-cli/coder"
  version  = "0.2.1"
  agent_id = coder_agent.main.id
  folder   = "/home/coder/project"

  # Optional
  install_cursor_cli = true
  force              = true
  model              = "gpt-5"
  ai_prompt          = data.coder_parameter.ai_prompt.value
  api_key            = "xxxx-xxxx-xxxx" # Required while using tasks, see note below

  # Minimal MCP server (writes `folder/.cursor/mcp.json`):
  mcp = jsonencode({
    mcpServers = {
      playwright = {
        command = "npx"
        args    = ["-y", "@playwright/mcp@latest", "--headless", "--isolated", "--no-sandbox"]
      }
      desktop-commander = {
        command = "npx"
        args    = ["-y", "@wonderwhy-er/desktop-commander"]
      }
    }
  })

  # Use a pre_install_script to install the CLI
  pre_install_script = <<-EOT
    #!/usr/bin/env bash
    set -euo pipefail
    curl -fsSL https://deb.nodesource.com/setup_20.x | bash -
    apt-get install -y nodejs
  EOT

  # Use post_install_script to wait for the repo to be ready
  post_install_script = <<-EOT
    #!/usr/bin/env bash
    set -euo pipefail
    TARGET="$${FOLDER}/.git/config"
    echo "[cursor-cli] waiting for $${TARGET}..."
    for i in $(seq 1 600); do
      [ -f "$TARGET" ] && { echo "ready"; exit 0; }
      sleep 1
    done
    echo "timeout waiting for $${TARGET}" >&2
  EOT

  # Provide a map of file name to content; files are written to `folder/.cursor/rules/<name>`.
  rules_files = {
    "python.mdc" = <<-EOT
        ---
        description: RPC Service boilerplate
        globs:
        alwaysApply: false
        ---

        - Use our internal RPC pattern when defining services
        - Always use snake_case for service names.
        
        @service-template.ts
      EOT

    "frontend.mdc" = <<-EOT
        ---
        description: RPC Service boilerplate
        globs:
        alwaysApply: false
        ---

        - Use our internal RPC pattern when defining services
        - Always use snake_case for service names.

        @service-template.ts
      EOT
  }
}

Note

A .cursor directory will be created in the specified folder, containing the MCP configuration, rules. To use this module with tasks, please pass the API Key obtained from Cursor to the api_key variable. To obtain the api key follow the instructions here

References

• See Cursor CLI docs: https://docs.cursor.com/en/cli/overview
• For MCP project config, see https://docs.cursor.com/en/context/mcp#using-mcp-json. This module writes your mcp_json into folder/.cursor/mcp.json.
• For Rules, see https://docs.cursor.com/en/context/rules#project-rules. Provide rules_files (map of file name to content) to populate folder/.cursor/rules/.

Troubleshooting

• Ensure the CLI is installed (enable install_cursor_cli = true or preinstall it in your image)
• Logs are written to ~/.cursor-cli-module/