Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Back-Agent MCP Server create a React component for a login form with validation"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
An MCP (Model Context Protocol) server that executes development tasks using Claude Code CLI.
Execute tasks through Claude Code CLI via MCP protocol
Non-interactive mode by default (-p flag auto-applied)
Specify custom working directories
Configurable timeout settings
Comprehensive error handling and logging
Related MCP server: Claude Code Connector MCP
Node.js >= 18
Claude Code CLI installed and available in PATH
# Clone the repository
git clone <repository-url>
cd back-agent-mcp
# Install dependencies
npm install
# Build the project
npm run build# Development mode (with tsx)
npm run dev
# Production mode (built)
npm startnpm install @zuens2020/back-agent-mcpAdd the following to your Claude Desktop configuration file:
Windows: %APPDATA%\Claude\claude_desktop_config.json
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Linux: ~/.config/Claude/claude_desktop_config.json
{
"mcpServers": {
"back-agent": {
"command": "node",
"args": ["--experimental-modules", "C:\\Users\\YourUsername\\AppData\\Roaming\\npm\\node_modules\\@zuens2020\\back-agent-mcp\\dist\\index.js"]
}
}
}Or using npx:
{
"mcpServers": {
"back-agent": {
"command": "npx",
"args": ["-y", "@zuens2020/back-agent-mcp"]
}
}
}Executes a development task using Claude Code CLI.
Parameters:
Name | Type | Required | Description |
| string | Yes | The task description to execute |
| string | No | Working directory for execution |
| number | No | Timeout in seconds (max 3600, default 300) |
| string[] | No | Additional CLI arguments (excluding |
Example:
{
"task": "Create a function that calculates fibonacci numbers",
"workingDirectory": "C:\\Projects\\my-app",
"timeout": 600
}# Type checking
npm run typecheck
# Build
npm run build
# Development mode
npm run devsrc/
├── index.ts # Main entry point
├── server/
│ └── tools/
│ └── execute-task.ts # Task execution tool
├── claude/
│ └── executor.ts # Claude Code CLI executor
└── utils/
├── logger.ts # Logging utilities
└── error-handler.ts # Error handlingVariable | Description | Values |
| Set logging verbosity |
|
MIT
Cancel a pending or running task.
Inputs: taskId (required).
IMPORTANT: This creates a background task for an AI programming assistant, NOT a direct shell command executor.
Spawns Claude Code (an AI coding agent) as a subprocess to complete development tasks in the background. Claude Code will interpret your natural language request and autonomously decide which actions to take.
What it does:
Creates a non-blocking background task
Returns task ID immediately for tracking
Up to 3 tasks run concurrently by default
What it does NOT do:
NOT a direct shell/bash command executor
Does NOT return raw stdout/stddr from commands
Returns a task ID. Use get-task-status to check progress and get-task-result to retrieve the output.
Inputs: task (required), workingDirectory, timeout, additionalArgs.
Delete a task from the task list. Use this to clean up old completed tasks.
Inputs: taskId (required).
IMPORTANT: This is an AI programming assistant tool, NOT a direct shell command executor.
This tool spawns Claude Code (an AI coding agent) as a subprocess to complete development tasks. Claude Code will interpret your natural language request and autonomously decide which actions to take (reading files, editing code, running commands, etc.) to accomplish the goal.
What it does:
Accepts natural language task descriptions
Claude Code AI figures out how to complete the task
Returns the AI's response and actions taken
What it does NOT do:
NOT a direct shell/bash command executor
Does NOT return raw stdout/stderr from commands
For direct command execution, use the Bash tool instead
Example usage:
"Create a REST API endpoint for user authentication"
"Debug why the tests are failing"
"Refactor the user module to use TypeScript"
The returned output is Claude Code's conversational response, not raw command output.
Inputs: task (required), workingDirectory, timeout, additionalArgs.
Get the full result of a task including stdout/stderr output. Only available for completed tasks.
Inputs: taskId (required).
Get statistics about all tasks.
Inputs: no named parameters.
Get the current status of a task. Returns status, timestamps, and basic info.
Inputs: taskId (required).
List all tasks, optionally filtered by status.
Inputs: status, limit.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 8 tools
There is significant overlap between create-task and execute-task, both described as spawning Claude Code AI for development tasks with similar capabilities and limitations, which could cause confusion. However, other tools like cancel-task, delete-task, get-task-result, get-task-status, list-tasks, and get-task-stats have clearer distinct purposes related to task management.
Tool names follow a consistent verb-noun pattern with hyphens (e.g., create-task, get-task-status), making them predictable and readable. The only minor deviation is execute-task, which uses a different verb than the others but still fits the pattern.
With 8 tools, the server is well-scoped for task management and AI programming assistance. Each tool has a clear role, such as creating, executing, monitoring, and managing tasks, which fits the server's purpose without being overly sparse or bloated.
The tool set covers the full lifecycle of tasks (create, execute, status, result, cancel, delete, list, stats), providing good coverage for task management. A minor gap is the lack of tools for updating or modifying tasks, but agents can work around this by recreating tasks if needed.
Work on your own computers from Claude or ChatGPT: run commands, edit and search files.
Source-checked CLI guides and model-aware planning for Claude Code, Codex, and Grok Build.
Create projects, nodes, and tasks in UluP Spaces by conversation with Claude.
Use your Mac, Windows or Linux computer from ChatGPT, Claude or Codex: files, commands, documents.
