home / mcp / github agentic chat mcp server

GitHub Agentic Chat MCP Server

Provides MCP access to search GitHub repos, create issues, and perform vector-based semantic searches.

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "akhidastech-github-agentic-chat-mcp": {
      "command": "/absolute/path/to/bin/github-agentic-chat-mcp",
      "args": [],
      "env": {
        "DATABASE_URL": "postgres://user:password@localhost:5432/dbname",
        "GITHUB_TOKEN": "YOUR_GITHUB_TOKEN",
        "OPENAI_API_KEY": "YOUR_OPENAI_API_KEY"
      }
    }
  }
}

You run a dedicated MCP server that allows you to interact with GitHub through natural language, perform repository searches, create issues, and leverage vector search for semantic retrieval. This server exposes friendly commands via an MCP client, enabling conversational access to GitHub data and actions without manual API juggling.

How to use

Use the MCP client to talk to the GitHub Agentic Chat MCP Server as if you were asking a knowledgeable assistant to search repositories, open issues, or retrieve relevant documents from your stored content. You can issue queries like” find repositories about CI pipelines,” “create an issue in owner/repo,” or “show me documents related to GitHub Actions.” The server integrates search capabilities with vector-based semantic search so you can retrieve contextually relevant results even when exact keywords aren’t used. Use it to streamline collaboration, automate routine GitHub tasks, and organize knowledge around your projects.

Typical workflows include: - Searching for repositories matching a topic and filtering by language or stars. - Creating issues directly from a chat prompt, with automatic assignment or labeling based on your instructions. - Adding documents to a vector store to enable semantic search over internal docs, guidelines, or design notes. - Running semantic searches to surface related documents and conversations that inform your current task.

How to install

Prerequisites you need before installing the MCP server: - Go 1.21 or later. - PostgreSQL with the pgvector extension installed. - A GitHub Personal Access Token with appropriate scopes. - An OpenAI API Key. - Claude Desktop or another MCP-compatible client to connect.

Step 1: Prepare your environment variables. You will configure tokens and access keys that the server uses at runtime.

export GITHUB_TOKEN=your_github_token_here
export DATABASE_URL=postgres://user:password@localhost:5432/dbname
export OPENAI_API_KEY=your_openai_api_key_here

Step 2: Set up PostgreSQL to support vector storage. You need the vector extension available in your database.

CREATE EXTENSION vector;

Step 3: Build the MCP server binary from source.

go build -o bin/github-agentic-chat-mcp

Step 4: Run the MCP server locally. The server runs as a local process and is configured via environment variables you set earlier.

bin/github-agentic-chat-mcp

Step 5: Connect a MCP client such as Claude Desktop. Use a configuration entry that points to the local server binary. For Claude Desktop, add a server entry with the absolute path to your built binary.

{
  "mcpServers": {
    "github-chat": {
      "command": "/absolute/path/to/bin/github-agentic-chat-mcp"
    }
  }
}

Additional notes

After starting the server, ensure Claude Desktop (or your MCP client) is configured to communicate with the local server. You can restart the client after starting the server to establish a new MCP connection.

Available tools

search_repositories

Search for GitHub repositories that match a given query string, with optional filters such as language, stars, and topic terms.

create_issue

Create a new issue in a specified GitHub repository, including title, body content, labels, and assignees.

add_to_vector_store

Add a document to the vector store, including content and metadata for semantic search.

vector_search

Perform a semantic search over stored documents using a query to retrieve contextually relevant results.