Mcptix

💡⚡️ Cost-Effective Development with mcptix

Having an API-connected LLM agent continuously planning, executing code, handling errors, and debugging can quickly become expensive. API costs add up when your agent needs to repeatedly process the same context and maintain state across interactions.

A more cost-effective approach:Configure Claude Desktop (with a Pro account running Claude 3.7 Sonnet) with mcptix and filesystem access. This local setup dramatically reduces API usage while maintaining powerful AI assistance.

By storing comprehensive plans in theagent_contextfield—complete with filenames, line numbers, method names, and other reference points organized into logical, workable chunks—you can reduce API usage by up to 80%.

mcptix acts as a next-generation memory bank for LLM coding agents, allowing them to offload detailed planning and context into a persistent storage system that they can reference as needed, rather than keeping everything in their limited context window.

What makes this system particularly powerful is that both your coding agent (like Roo/Cline) and Claude Desktop access the same underlying mcptix database. This creates a seamless collaborative environment where Claude Desktop can create comprehensive plans and store them in tickets, while your coding agent can retrieve these tickets and execute the plans precisely. They effectively communicate through the shared ticket system—Claude Desktop breaking down complex tasks into executable chunks, and your coding agent implementing them without needing to regenerate the context each time.

Configuration for Different AI Assistants

When you runnpx mcptix init, an MCP server configuration file is created at.mcptix/mcp-server-config.json. Copy this to Roo's configuration directory:

mkdir -p .roo cp .mcptix/mcp-server-config.json .roo/mcp.json

Check the configuration fileto ensure paths are absolute:

{ "mcpServers": { "mcptix": { "command": "/absolute/path/to/node", "args": ](#ai-integration-guide)["/absolute/path/to/node_modules/@ownlytics/mcptix/dist/mcp/index.js"], "env": { "MCPTIX_HOME_DIR": "/absolute/path/to/your/project/.mcptix", "HOME": "/home/your-username" }, "disabled": false, "alwaysAllow": [] } } }

Ensure/absolute/path/to/nodeis the result of runningwhich nodein your terminal.

Ensure/absolute/path/to/node_modules/@ownlytics/mcptix/dist/mcp/index.jsis the absolute path to the mcptix MCP server in your node_modules.

Ensure/absolute/path/to/your/project/.mcptixis the absolute path to your project's.mcptixdirectory.

Ensure/home/your-usernameis your home directory (result ofecho $HOME).
-

Install desktop-commanderif you haven't already:

npm install -g @wonderwhy-er/desktop-commander

Connect Claude Desktop to your filesystem:

- Go to Settings > Developer
- Add the MCP configuration in the "MCP Server Configuration" section:

{ "mcpServers": { "mcptix": { "command": "/absolute/path/to/node", "args": ["/absolute/path/to/node_modules/@ownlytics/mcptix/dist/mcp/index.js"], "env": { "MCPTIX_HOME_DIR": "/absolute/path/to/your/project/.mcptix", "HOME": "/home/your-username" }, "disabled": false, "alwaysAllow": [] }, "desktop-commander": { "command": "npx", "args": ["-y", "@smithery/cli@latest", "run", "@wonderwhy-er/desktop-commander", "--config", "{}"] } } }

Use the json as described in the coding agent configuration.

- Click "New Project" in Claude Desktop
- Name your project appropriately
- Set your project directory to your development project's root directory

Add instructions for Claude about using mcptix:

This project uses mcptix for ticket tracking. When planning work: 1. Use the mcptix MCP server to create and manage tickets 2. Break down complex tasks into smaller tickets 3. Use the agent_context field in tickets to store comprehensive planning in Markdown format 4. Project files are located in: /path/to/your/project

Once configured, you can start talking to your AI assistant about your project. Here's a workflow:

Initiate a conversationwith Claude Desktop or your AI assistant about planning your project

Ask for help breaking down complex tasks: Example: "I need to implement a user authentication system. Can you help me break this down into manageable tickets?"

The AI will create ticketswith detailed planning in theagent_contextfield

View the ticketsin the mcptix UI athttp://localhost:3000(or your configured port)

Find tickets in the 'in-progress' status.

Work on ticket id ticket-1234567890 following the plan in the agent_context.

module.exports = { // Database configuration dbPath: './.mcptix/data/mcptix.db', // API server configuration apiPort: 3000, apiHost: 'localhost', // Server options mcpEnabled: false, // Disabled by default - MCP server should be started by the LLM agent apiEnabled: true, // Logging configuration logLevel: 'info', // Data management clearDataOnInit: false, };

# Initialize mcptix in your project npx mcptix init # Start mcptix npx mcptix start # Start with custom port and host npx mcptix start --port 3001 --host 0.0.0.0 # Start without opening the browser npx mcptix start --no-open # Start only the MCP server (for development/testing purposes) npx mcptix mcp

📧hello@ownlytics.io
📄View full license terms

Interact with task, doc, and project data in Dart, an AI-native project management tool