This is a demonstration of using AIGNE Framework and GitHub MCP Server to interact with GitHub repositories.
flowchart LR
in(In)
out(Out)
agent(AI Agent)
github(GitHub MCP Agent)
searchRepos(Search Repositories)
getContents(Get Contents)
createFile(Create/Update File)
issue(Issues & PRs)
in --> agent <--> github
subgraph MCP Agent
github <--> searchRepos
github <--> getContents
github <--> createFile
github <--> issue
end
agent --> out
classDef inputOutput fill:#f9f0ed,stroke:#debbae,stroke-width:2px,color:#b35b39,font-weight:bolder;
classDef processing fill:#F0F4EB,stroke:#C2D7A7,stroke-width:2px,color:#6B8F3C,font-weight:bolder;
class in inputOutput
class out inputOutput
class agent processing
class github processing
class searchRepos processing
class getContents processing
class createFile processing
class issue processing
Following is a sequence diagram of the workflow to search for repositories and access contents:
sequenceDiagram
participant User
participant AI as AI Agent
participant G as GitHub MCP Agent
participant SR as Search Repositories
participant GC as Get Contents
User ->> AI: Search for repositories related to 'modelcontextprotocol'
AI ->> G: Use search_repositories function
G ->> SR: Execute search with query='modelcontextprotocol'
SR ->> G: Return repository list results
G ->> AI: Formatted repository information
AI ->> User: Found these repositories: modelcontextprotocol/servers, etc.
User ->> AI: Get README from the servers repo
AI ->> G: Use get_file_contents function
G ->> GC: Access README.md in modelcontextprotocol/servers
GC ->> G: Return file content
G ->> AI: Formatted file content
AI ->> User: Here's the README content: ...
- Node.js and npm installed on your machine
- An OpenAI API key for interacting with OpenAI's services
- GitHub Personal Access Token with appropriate permissions
- Optional dependencies (if running the example from source code):
export OPENAI_API_KEY=YOUR_OPENAI_API_KEY # Set your OpenAI API key
export GITHUB_TOKEN=YOUR_GITHUB_TOKEN # Set your GitHub token
npx -y @aigne/example-mcp-github # Run the example
git clone https://github.com/AIGNE-io/aigne-framework
cd aigne-framework/examples/mcp-github
pnpm install
Setup your API keys in the .env.local
file:
OPENAI_API_KEY="" # Set your OpenAI API key here
GITHUB_TOKEN="" # Set your GitHub Personal Access Token here
pnpm start # Interactive chat interface
# or
pnpm example # Run predefined examples
The following example demonstrates how to use the GitHub MCP server to search for repositories:
import { AIAgent, AIGNE, MCPAgent } from "@aigne/core";
import { OpenAIChatModel } from "@aigne/core/models/openai-chat-model.js";
// Load environment variables
const { OPENAI_API_KEY, GITHUB_TOKEN } = process.env;
// Initialize OpenAI model
const model = new OpenAIChatModel({
apiKey: OPENAI_API_KEY,
});
// Initialize GitHub MCP agent
const githubMCPAgent = await MCPAgent.from({
command: "npx",
args: ["-y", "@modelcontextprotocol/server-github"],
env: {
GITHUB_TOKEN,
},
});
// Create AIGNE
const aigne = new AIGNE({
model,
skills: [githubMCPAgent],
});
// Create AI agent with GitHub-specific instructions
const agent = AIAgent.from({
instructions: `\
## GitHub Interaction Assistant
You are an assistant that helps users interact with GitHub repositories.
You can perform various GitHub operations like:
1. Searching repositories
2. Getting file contents
3. Creating or updating files
4. Creating issues and pull requests
5. And many more GitHub operations
Always provide clear, concise responses with relevant information from GitHub.
`,
});
// Example: Search for repositories
const result = await aigne.invoke(
agent,
"Search for repositories related to 'modelcontextprotocol'"
);
console.log(result);
// Output:
// I found several repositories related to 'modelcontextprotocol':
//
// 1. **modelcontextprotocol/servers** - MCP servers for various APIs and services
// 2. **modelcontextprotocol/modelcontextprotocol** - The main ModelContextProtocol repository
// ...
// Shutdown the aigne when done
await aigne.shutdown();
The GitHub MCP server provides a wide range of operations including:
-
Repository Operations:
- Search repositories
- Create repositories
- Get repository information
-
File Operations:
- Get file contents
- Create or update files
- Push multiple files in a single commit
-
Issue and PR Operations:
- Create issues
- Create pull requests
- Add comments
- Merge pull requests
-
Search Operations:
- Search code
- Search issues
- Search users
-
Commit Operations:
- List commits
- Get commit details
This project is licensed under the MIT License.