agent-pm
MCP server for the planning and execution of AI-assisted development projects.
claude mcp add --transport stdio gannonh-agent-pm npx -y @gannonh/agent-pm@latest \ --env PROJECT_ROOT="/path/to/project/root/" \ --env ANTHROPIC_API_KEY="sk-your-anthropic-api-key" \ --env PERPLEXITY_API_KEY="pplx-your-perplexity-api-key"
How to use
AgentPM is a planning and orchestration system that runs locally as an MCP server to help you manage AI-powered software development projects. It acts as a product manager within your IDE or coding environment, guiding you to develop requirements, decompose work into actionable tasks with dependencies, and orchestrate implementation with context-aware assistance. It integrates with Claude Sonnet 3.7 for task generation and can optionally leverage Perplexity for research-backed insights, all while producing well-structured documentation and living project context via Context7 integration. The toolset is designed to minimize context-switching and memory management by delivering the right information at the right time and returning outputs in clear, human-readable markdown when needed.
To use AgentPM, ensure you have the required API keys and a supported IDE that implements the Model Context Protocol. When running the MCP server, you can interact with it through your IDE’s MCP integration (Cursor, Augment, VS Code Copilot, Cline, Roo, etc.). AgentPM will generate and manage tasks, track priorities, and provide implementation details and dependencies, making it easier to plan and execute complex software development efforts within your existing workflow.
How to install
Prerequisites
- Node.js version 20.0.0 or higher
- Anthropic API Key (for Claude AI integration)
- Perplexity API Key (for enhanced research-backed task generation)
Installation steps
- Install the MCP server using npx (as shown in the README):
# Ensure Node.js is installed and up to date
node -v
npm -v
# Start AgentPM via npx (no global install required)
npx -y @gannonh/agent-pm@latest
- Configure environment variables
- Create or edit your MCP configuration (Cursor/mcp.json or Augment settings as shown in the README).
- Add required environment variables:
- PROJECT_ROOT: Path to your project root
- ANTHROPIC_API_KEY: Your Claude API key
- PERPLEXITY_API_KEY: Your Perplexity API key (optional but recommended)
- Optional IDE integration
- For Cursor, place the mcpServers block in
.cursor/mcp.jsonas demonstrated in the README. - For Augment, add the corresponding mcpServers configuration to your VS Code settings.json under augment.advanced.
- Verify configuration
- Ensure the MCP server block appears in your IDE’s MCP configuration and that environment variables are correctly set. Start the agent from your IDE and confirm that it responds to prompts and generates tasks and documentation as expected.
Additional notes
Tips and common considerations:
- Keep ANTHROPIC_API_KEY and PERPLEXITY_API_KEY secure; avoid committing them to version control.
- If you encounter token or memory issues, adjust optional environment variables like DEBUG_LOGS, ANTHROPIC_MAX_TOKENS, or PERPLEXITY related settings according to your project's needs.
- The agent can generate structured markdown documentation automatically; you can use this output directly in your project docs.
- Ensure your project root path (PROJECT_ROOT) is accurate to enable correct context management and file references.
- If you switch IDEs, update the MCP configuration accordingly (Cursor, Augment, Copilot, Cline, etc.).
- Monitor usage and costs of Claude and Perplexity APIs to manage API quotas effectively.
Related MCP Servers
BifrostMCP
VSCode Extension with an MCP server that exposes semantic tools like Find Usages and Rename to LLMs
ls
List MCP Server configurations in your system used by AI applications like Cursor, Claude Desktop, VS Code and others
kanban
MCP Kanban is a specialized middleware designed to facilitate interaction between Large Language Models (LLMs) and Planka, a Kanban board application. It serves as an intermediary layer that provides LLMs with a simplified and enhanced API to interact with Planka's task management system.
adb
An MCP (Model Context Protocol) server for interacting with Android devices through ADB in TypeScript.
mcp-image-extractor
MCP server which allow LLM in agent mode to analyze image whenever it needs
memory
A MCP (Model Context Protocol) server providing long-term memory for LLMs