ython environment - Maintain proper context isolation between projects ### Traditional CLI Mode ```bash python main.py --mode cli ``` Available commands: #### Memory - `/memory add ` - Add a new general memory - `/memory get ` - Search existing memories - `/memory repo add ` - Add a repository-specific memory - `/memory repo get ` - Search repository-specific memories #### Git and GitHub - `/git commits [number]` - Show commit history (default: last 5) - `/git issues` - List local repository issues - `/git info` - Show detailed repository information - `/git diff` - Show pending changes (staged and unstaged) - `/github repo ` - Show repository details - `/github issues [state]` - List issues (state: open/closed) - `/github prs [state]` - List pull requests (state: open/closed) - `/github project ` - Show project information - `/github summarize ` - Generate issue summary using GPT - `/github search [language]` - Search code on GitHub #### Code Analysis - `/code analyze [language]` - Analyze code structure and provides: - High-level overview of the code's purpose - Key components and their responsibilities - Notable patterns or techniques used - Potential improvements and best practices - Security considerations when relevant - Structural analysis (functions, classes, imports) Example: ```bash mcp-dev-agent-cli > /code analyze src/main.py ``` #### Documentation - `/docs ` - Search documentation - `exit` - Exit CLI mode ### Advanced Features #### Code Analysis The system uses tree-sitter to analyze code in: - Python - JavaScript - TypeScript Analysis includes: - Function and method extraction - Class identification - Import mapping - Memory integration for context #### GitHub Integration - Detailed repository information - Issue search and listing - Pull request management - GitHub Projects integration - Code search with language filtering - Enriched metadata - Automatic issue summarization with GPT #### Enhanced Memory System - Support for different context types - Enriched metadata - Contextual search - Temporal history - Code analysis integration - Automatic summary storage - Git state contextualization ### Server Mode (SSE) ```bash python main.py # or python main.py --mode server ``` The SSE server enables integration with other applications through the MCP protocol. ## Project Structure ``` . ├── main.py # Application entry point ├── setup_parsers.py # Code parser configuration ├── tools/ │ ├── memory_tool.py # Memory management via ChromaDB │ ├── doc_tool.py # Documentation search │ ├── git_tool.py # Git integration │ └── github_tool.py # GitHub integration and code analysis └── docs/ └── api_reference.md # API reference documentation ``` ## AI Model Configuration The agent uses Ollama's local models for AI features. The following models are configured: 1. CodeLlama (default) - Optimized for code-related tasks 2. Llama 2 - General purpose tasks 3. Mistral - Alternative model for various tasks The `.agent.json` file is already configured with: - CodeLlama as default model - Response caching (24 hours) - Interaction logging ### Model Parameters You can adjust model parameters by editing the `.agent.json` file: - `temperature`: Controls response creativity (0.0 to 1.0) - `context_window`: Maximum context length - `top_p`: Controls response diversity ### Using Different Models Different features use specialized models: - Code documentation: CodeLlama - Issue summarization: CodeLlama - General queries: Llama 2 No API keys required - all AI features run locally through Ollama!2a:T6eb,## What is MCP Development Agent? MCP Development Agent is a development tool based on the Model Context Protocol (MCP) that offers features for memory management, documentation search, and Git/GitHub integration, available as both a CLI tool and an SSE serv

MCP Development Agent

Comments