The Narad GitHub Agent is a next-generation, dual-mode intelligent agent engineered for seamless high-level interactions with GitHub. It operates as an interactive CLI chat agent you can talk to directly, or as a Model Context Protocol (MCP) server that integrates with AI-powered IDEs and tools.
Powered by Gemini 2.0 Flash, it transforms complex repository tasks into simple conversational interactions β with conversation memory, AI code reviews, and a daily digest of your GitHub activity.
Fully Dynamic: The agent uses whoever's
GITHUB_TOKENis in.env. Clone it, put your own token in, and it works 100% for your account β no code changes needed.
| Feature | Description |
|---|---|
| π¬ Interactive CLI Mode | Chat with the agent directly in your terminal using natural language |
| π MCP Server Mode | Integrate with AI IDEs (VS Code + Antigravity, Claude Desktop) via MCP |
| π§ Conversation Memory | Every session is stored in a local SQLite DB β Gemini remembers context |
| π₯ PR Code Reviewer | Fetch a PR's diff and get an expert AI code review (bugs, security, rating) |
| π Daily Digest | AI-written morning report of activity across all your repos |
| π€ Profile Overview | View any GitHub user's bio, stats, and top repos with an AI Developer Summary |
| β‘ Gemini 2.0 Flash | Reasoning engine for repo health, code analysis, and GitHub Q&A |
| π Security First | .env is git-ignored. Your token is never pushed to GitHub |
| π¦ Clean Architecture | Pydantic Settings, structured logging, modular tool design |
| Technology | Role |
|---|---|
| Gemini 2.0 Flash | AI Reasoning & Analysis Engine |
| FastMCP | Model Context Protocol Server |
| PyGitHub | GitHub REST API Integration |
| SQLite (stdlib) | Local DB for conversation memory, PR reviews & digests |
| Pydantic Settings | Config Management & Validation |
| Rich | Beautiful Terminal UI/CLI |
| Python 3.10+ | Core Language |
| venv | Isolated Dependency Management |
Narad-GitHub-Agent/
βββ src/
β βββ main.py # Dual entry point: CLI or MCP Server
β βββ narad_mcp/
β βββ config.py # Pydantic Settings & Config
β βββ server.py # FastMCP Server with all tools
β βββ cli.py # Rich Interactive CLI Agent
β βββ database.py # SQLite layer (memory, PR reviews, digests)
β βββ tools/
β β βββ github_tools.py # All GitHub API interactions
β βββ agents/
β βββ gemini_agent.py # Gemini 2.0 Flash Reasoning Engine
βββ tests/ # Quality assurance suite
βββ narad_agent.db # π Local SQLite DB (auto-created, git-ignored)
βββ .env # π Secrets (git-ignored, never pushed)
βββ .env.example # Credential Template for setup
βββ .gitignore # Strict security rules
βββ requirements.txt # Modern dependencies
Ensure you have Python 3.10+ installed and your API keys ready.
# Clone the repo
git clone https://github.com/Edge-Explorer/Narad-GitHub-Agent.git
cd Narad-GitHub-Agent
# Create and activate virtual environment
python -m venv venv
.\venv\Scripts\activate # Windows
# Install dependencies
pip install -r requirements.txtCreate a .env file by copying the example:
copy .env.example .envThen fill in your keys:
GITHUB_TOKEN="your_github_personal_access_token"
GEMINI_API_KEY="your_gemini_api_key"Getting a GitHub Token: Go to GitHub Settings β Developer Settings β Personal Access Tokens and create a token with
reposcope.Getting a Gemini API Key: Visit Google AI Studio to generate your free Gemini API key.
python src/main.py --clipython src/main.pyYou can type natural language or use the keywords below:
| Command | What it does |
|---|---|
repos |
List your repositories (auto-detected from token) |
repos <username> |
List any GitHub user's public repos |
profile |
Your full GitHub profile: bio, stats, top repos |
profile <username> |
Any user's full profile & top repos |
commits <owner/repo> |
Recent commit history for a repo |
branches <owner/repo> |
List all branches |
prs <owner/repo> |
List open Pull Requests |
review pr <owner/repo> <#> |
π₯ AI Code Review of a specific PR |
analyze <owner/repo> |
π§ AI repo health report + score out of 10 |
digest |
π Generate today's Daily GitHub Digest |
history |
View your current session's conversation history |
ask <question> |
Ask Gemini any GitHub or coding question |
exit |
Quit the agent |
show me my repos
give me commits from Edge-Explorer/Narad-GitHub-Agent
analyze my Narad project
what are the open PRs in Edge-Explorer/Narad-MCP-Server
review pr Edge-Explorer/my-project 3
give me today's digest
what is a pull request?
The agent automatically creates narad_agent.db to persist data locally:
| Table | Stores |
|---|---|
conversation_history |
All questions & answers per session (with session ID) |
pr_reviews |
Full AI code reviews with repo, PR number & timestamp |
daily_digests |
All generated daily digests, so you can compare over time |
The database file is git-ignored and stays only on your machine.
- Your token is 100% dynamic: The agent reads
GITHUB_TOKENfrom.envat startup. If you change the token, it uses the new one automatically. No code changes needed. - Token never leaks:
.envand*.dbare strictly in.gitignoreβ they will never be pushed to GitHub. - Anyone can clone this: A new user clones the repo, adds their own
.envwith their credentials, and the agent will work for their GitHub account β not yours. - Least Privilege: Only the GitHub token scopes you need (
repois enough for all features). - Validated Config: Pydantic Settings validates all environment variables on startup β it fails fast with a clear error if a key is missing.
When running as an MCP Server, these tools are exposed to your AI IDE:
| Tool | Description |
|---|---|
list_repositories |
List repos for you or any GitHub user |
get_commits |
Fetch recent commit history |
analyze_repository |
AI-powered health & activity report |
read_file |
Read any file from any branch |
search_github_code |
Search code across all of GitHub |
list_repo_branches |
List all branches |
ask_gemini_github |
Ask Gemini any GitHub/dev question |
Narad GitHub Agent β Re-engineered from the ground up with a focus on security, modularity, AI-first design, and genuine day-to-day developer utility.
Built as part of the Narad AI Ecosystem β a family of modular, premium AI agents.
This project is licensed under the MIT License. See the LICENSE file for details.
Β© 2026 Edge-Explorer Β· Narad AI Ecosystem