Your Personal AI Development Assistant powered by the Model Context Protocol (MCP).

DevPilot MCP is a local AI-powered developer assistant that allows LLMs such as ChatGPT, Claude, or any MCP-compatible client to securely interact with your development environment.

Instead of manually running Git commands, searching files, checking Docker containers, or querying databases, DevPilot exposes these capabilities as MCP tools.

The long-term vision is to create an intelligent software engineering assistant that can understand natural language, invoke the correct tools, and automate everyday development workflows.

---

✨ Features

✅ Git Tools

• Git Status // done
• Current Branch //
• Commit History //
• Diff //
• Show Commit
• Checkout Branch
• Create Branch
• Commit Changes
• Push / Pull
• Stash Operations

---

📁 Filesystem Tools

• Read File
• Write File
• Search Files
• List Directory
• Find TODOs
• Rename Files
• Delete Files

---

🐳 Docker Tools

• List Containers
• View Logs
• Restart Containers
• Stop Containers
• List Images

---

🗄 Database Tools

• Execute SQL Queries
• Show Schema
• Describe Tables
• List Databases
• Connection Information

---

🌐 GitHub Tools

• Clone Repository
• Create Pull Request
• Create Issue
• List Pull Requests
• List Branches

---

🤖 AI Assistant (Upcoming)

Instead of

Choose Tool:
git_status

You'll simply ask

Show me the status of my data-api repository.

The assistant will

• Understand your request
• Choose the correct MCP tool
• Execute it
• Explain the result naturally

---

🧩 Planner Arg Extractors

Defined in client/agent/planner.py as ARG_EXTRACTORS.

Each key is a tool argument name. The value is a function that pulls that argument out of the user's natural-language prompt (or None until implemented).

Key	Meaning
repo_path	Absolute or relative path to a Git repository
path	Filesystem path (file or directory)
symbol	Code symbol name (class, function, etc.)
branch	Git branch name
container	Docker container name or ID
command	Shell command to run (e.g. inside a container)
commit	Git commit hash or ref (HEAD, abc123, …)
content	Text content to write into a file
message	Commit message or stash message
query	Text search string
pattern	Glob pattern for finding files
remote	Git remote name (origin, upstream, …)
operation	Stash operation (list, push, pop, apply, drop)
limit	Max number of items (e.g. git log entries)
start_line	First line number when reading a file
max_lines	How many lines to read from a file
add_all	Whether to git add -A before commit (true/false)
append	Whether to append instead of overwrite a file (true/false)
checkout	Whether to check out a branch after creating it (true/false)
staged	Whether to show staged diffs only (true/false)

Example shape:

ARG_EXTRACTORS = {
    "repo_path": extract_repo_path,  # implemented
    "branch": None,                 # TODO: implement later
    ...
}

---

🏗 Architecture

                   User Prompt
                       │
                       ▼
                 Planner (Brain)
                       │
        ┌──────────────┼──────────────┐
        ▼              ▼              ▼
    Git Tools      FS Tools      Docker Tools
        │              │              │
        └──────────────┼──────────────┘
                       ▼
                Collect Results
                       │
                       ▼
                LLM Final Answer

---

# 📂 Project Structure

devpilot-mcp/ │ ├── README.md ├── pyproject.toml ├── .env │ ├── client/ │ ├── client.py │ ├── llm.py │ ├── prompt.py │ └── ui.py │ ├── server/ │ ├── server.py │ │ │ ├── git/ │ ├── filesystem/ │ ├── docker/ │ ├── postgres/ │ └── utils/ │ └── tests/

---

# �� Current Progress

## Phase 1 — MCP Fundamentals

- [x] MCP Server
- [x] MCP Client
- [x] Session Initialization
- [x] Tool Discovery
- [x] Tool Invocation

---

## Phase 2 — Git Tools

- [x] hello()
- [x] git_status()
- [x] git_branch()

### Upcoming

- [ ] git_log()
- [ ] git_diff()
- [ ] git_show()
- [ ] git_add()
- [ ] git_commit()
- [ ] git_push()

---

## Phase 3 — Filesystem

- [ ] read_file()
- [ ] write_file()
- [ ] search_files()
- [ ] list_directory()

---

## Phase 4 — Docker

- [ ] docker_ps()
- [ ] docker_logs()
- [ ] docker_restart()

---

## Phase 5 — PostgreSQL

- [ ] run_query()
- [ ] describe_table()
- [ ] list_tables()

---

## Phase 6 — GitHub

- [ ] create_pr()
- [ ] create_issue()
- [ ] clone_repo()

---

## Phase 7 — AI Assistant

- [ ] Natural language interface
- [ ] Automatic tool selection
- [ ] Context awareness
- [ ] Conversation memory

---

# 🛠 Installation

Clone the repository

```bash
git clone https://github.com/<your-username>/devpilot-mcp.git
cd devpilot-mcp

Install Poetry dependencies

poetry install

Activate the virtual environment

poetry shell

Run the server

poetry run python server/server.py

Run the client

poetry run python client/client.py

---

📖 Example

User

Show me my current Git branch.

Assistant

Calling git_branch(...)

Output

feature/devpilot

---

🎯 Vision

The goal of DevPilot MCP is to become a complete AI-powered development companion capable of understanding natural language and safely interacting with your local development environment.

Instead of memorizing terminal commands, developers should be able to simply describe what they want to accomplish.

Examples:

• "Show me modified files."
• "Create a new branch called feature/auth."
• "Find all TODO comments."
• "Restart my Docker containers."
• "Run the database migration."
• "Explain the last five commits."

DevPilot will determine which tools to use, execute them, and present the results in a human-friendly way.

---

📚 Tech Stack

• Python
• FastMCP
• Model Context Protocol (MCP)
• AsyncIO
• Poetry
• Git
• Docker
• PostgreSQL
• OpenAI / Claude (Planned)

---

🤝 Contributing

Contributions, issues, and feature requests are welcome.

Feel free to open an issue or submit a pull request.

---

📜 License

MIT License

---

⭐ Roadmap

• MCP Fundamentals
• Git Automation
• Filesystem Automation
• Docker Integration
• PostgreSQL Integration
• GitHub Automation
• AI Tool Selection
• Memory
• Multi-Agent Support
• Voice Commands
• VS Code Extension