agentic-tools

Agentic Tools MCP Server provides an integrated environment for managing AI-driven tasks and agent memories within project contexts. It combines advanced task management with an agent memories system, enabling persistent storage, multi-field search, and memory-based context to enhance AI assistants. The server exposes a suite of MCP tools for projects, tasks, memories, and AI-assisted features like PRD parsing, task recommendations, research queries, and complexity analysis, all designed to work with a per-project workingDirectory unless you opt into a global Claude-compatible workspace. Tools are accessed via the command line using the MCP server in your project directory or through integrations (e.g., Claude or AugmentCode) that reference the MCP server.

To use the server, start it through npx (or your preferred installation method) and then invoke the available tools by name with a specified workingDirectory. For project-specific workflows, data is isolated per project so you can manage tasks, memories, and project data independently. When using Claude Desktop with the global workspace, you can consolidate data across projects by enabling the Claude mode, which switches the storage to a shared global directory and reports the workingDirectory parameter as ignored in favor of the global path.

Prerequisites:

• Node.js and npm or pnpm installed on your system
• Internet access to install packages from npm

Option A: Quick Start (no global install required)

npx -y @pimzino/agentic-tools-mcp

Option B: Global installation (preferred for CLI usability)

npm install -g @pimzino/agentic-tools-mcp

Option C: Use via npm script in your project (alternative for local usage)

npm install --save-dev @pimzino/agentic-tools-mcp

After installing, you can run the MCP server in your project directory and start using the provided tools to manage projects, tasks, memories, and AI-assisted workflows. If you plan to use Claude integration, refer to the Claude-specific run modes described in the usage guide.

Notes and tips:

• The MCP server supports two storage modes: project-specific (default) and global Claude directory mode. Use the --claude flag to switch to the global directory for cross-project work.
• When using the Claude global workspace, the workingDirectory parameter is ignored and the data is stored in a standardized global location.
• All MCP tools require a workingDirectory parameter to specify where to store data. This enables per-project task and memory management.
• The suite includes powerful AI-enabled tools such as parse_prd, get_next_task_recommendation, analyze_task_complexity, infer_task_progress, research_task, and generate_research_queries for enhanced automation and insight generation.
• If you’re integrating with the VS Code extension or AugmentCode, ensure you configure the MCP server with the appropriate command and arguments as shown in the usage examples.
• When upgrading from older task models, the migrate_subtasks tool can help convert legacy subtasks to the unified model with unlimited depth.