Vercel AI SDK Documentation MCP Agent

1# Vercel AI SDK Documentation MCP Agent
2 
3A Model Context Protocol (MCP) server that provides AI-powered search and querying capabilities for the Vercel AI SDK documentation. This project enables developers to ask questions about the Vercel AI SDK and receive accurate, contextualized responses based on the official documentation.
4 
5[![MCP Compatible](https://img.shields.io/badge/MCP-Compatible-green)](https://modelcontextprotocol.io)
6[![TypeScript](https://img.shields.io/badge/TypeScript-4.9+-blue.svg)](https://www.typescriptlang.org/)
7[![Node.js](https://img.shields.io/badge/Node.js-18+-green.svg)](https://nodejs.org/)
8 
9## Features
10 
11- **Direct Documentation Search**: Query the Vercel AI SDK documentation index directly using similarity search
12- **AI-Powered Agent**: Ask natural language questions about the Vercel AI SDK and receive comprehensive answers
13- **Session Management**: Maintain conversation context across multiple queries
14- **Automated Indexing**: Includes tools to fetch, process, and index the latest Vercel AI SDK documentation
15 
16## Architecture
17 
18This system consists of several key components:
19 
201. **MCP Server**: Exposes tools via the Model Context Protocol for integration with AI assistants
212. **DocumentFetcher**: Crawls and processes the Vercel AI SDK documentation
223. **VectorStoreManager**: Creates and manages the FAISS vector index for semantic search
234. **AgentService**: Provides AI-powered answers to questions using the Google Gemini model
245. **DirectQueryService**: Offers direct semantic search of the documentation
25 
26## Setup Instructions
27 
28### Prerequisites
29 
30- Node.js 18+
31- npm
32- A Google API key for Gemini model access
33 
34### Environment Variables
35 
36Create a `.env` file in the project root with the following variables:
37 
38```
39GOOGLE_GENERATIVE_AI_API_KEY=your-google-api-key-here
40```
41 
42You'll need to obtain a Google Gemini API key from the [Google AI Studio](https://makersuite.google.com/app/apikey).
43 
44### Installation
45 
461. Clone the repository
47   ```
48   git clone https://github.com/IvanAmador/vercel-ai-docs-mcp.git
49   cd vercel-ai-docs-mcp-agent
50   ```
51 
522. Install dependencies
53   ```
54   npm install
55   ```
56 
573. Build the project
58   ```
59   npm run build
60   ```
61 
624. Build the documentation index
63   ```
64   npm run build:index
65   ```
66 
675. Start the MCP server
68   ```
69   npm run start
70   ```
71 
72## Integration with Claude Desktop
73 
74[Claude Desktop](https://www.anthropic.com/claude/download) is a powerful AI assistant that supports MCP servers. To connect the Vercel AI SDK Documentation MCP agent with Claude Desktop:
75 
761. First, install [Claude Desktop](https://www.anthropic.com/claude/download) if you don't have it already.
77 
782. Open Claude Desktop settings (via the application menu, not within the chat interface).
79 
803. Navigate to the "Developer" tab and click "Edit Config".
81 
824. Add the Vercel AI Docs MCP server to your configuration:
83 
84```json
85{
86  "mcpServers": {
87    "vercel-ai-docs": {
88      "command": "node",  
89      "args": ["ABSOLUTE_PATH_TO_PROJECT/dist/main.js"],
90      "env": {
91        "GOOGLE_GENERATIVE_AI_API_KEY": "your-google-api-key-here"
92      }
93    }
94  }
95}
96```
97 
98Make sure to replace:
99- `ABSOLUTE_PATH_TO_PROJECT` with the actual path to your project folder
100- `your-google-api-key-here` with your Google Gemini API key
101 
1025. Save the config file and restart Claude Desktop.
103 
1046. To verify the server is connected, look for the hammer 🔨 icon in the Claude chat interface.
105 
106For more detailed information about setting up MCP servers with Claude Desktop, visit the [MCP Quickstart Guide](https://modelcontextprotocol.io/quickstart/user).
107 
108## Integration with Other MCP Clients
109 
110This MCP server is compatible with any client that implements the Model Context Protocol. Here are a few examples:
111 
112### Cursor
113 
114[Cursor](https://cursor.sh/) is an AI-powered code editor that supports MCP servers. To integrate with Cursor:
115 
1161. Add a `.cursor/mcp.json` file to your project directory (for project-specific configuration) or a `~/.cursor/mcp.json` file in your home directory (for global configuration).
117 
1182. Add the following to your configuration file:
119 
120```json
121{
122  "mcpServers": {
123    "vercel-ai-docs": {
124      "command": "node",  
125      "args": ["ABSOLUTE_PATH_TO_PROJECT/dist/main.js"],
126      "env": {
127        "GOOGLE_GENERATIVE_AI_API_KEY": "your-google-api-key-here"
128      }
129    }
130  }
131}
132```
133 
134For more information about using MCP with Cursor, refer to the [Cursor MCP documentation](https://modelcontextprotocol.io/example-clients/).
135 
136## Usage
137 
138The MCP server exposes three primary tools:
139 
140### 1. agent-query
141 
142Query the Vercel AI SDK documentation using an AI agent that can search and synthesize information.
143 
144```json
145{
146  "name": "agent-query",
147  "arguments": {
148    "query": "How do I use the streamText function?",
149    "sessionId": "unique-session-id"
150  }
151}
152```
153 
154### 2. direct-query
155 
156Perform a direct similarity search against the Vercel AI SDK documentation index.
157 
158```json
159{
160  "name": "direct-query",
161  "arguments": {
162    "query": "streamText usage",
163    "limit": 5
164  }
165}
166```
167 
168### 3. clear-memory
169 
170Clears the conversation memory for a specific session or all sessions.
171 
172```json
173{
174  "name": "clear-memory",
175  "arguments": {
176    "sessionId": "unique-session-id"
177  }
178}
179```
180 
181To clear all sessions, omit the sessionId parameter.
182 
183## Development
184 
185### Project Structure
186 
187```
188├── config/              # Configuration settings
189├── core/                # Core functionality
190│   ├── indexing/        # Document indexing and vector store
191│   └── query/           # Query services (agent and direct)
192├── files/               # Storage directories
193│   ├── docs/            # Processed documentation
194│   ├── faiss_index/     # Vector index files
195│   └── sessions/        # Session data
196├── mcp/                 # MCP server and tools
197│   ├── server.ts        # MCP server implementation
198│   └── tools/           # MCP tool definitions
199├── scripts/             # Build and utility scripts
200└── utils/               # Helper utilities
201```
202 
203### Build Scripts
204 
205- `npm run build`: Compile TypeScript files
206- `npm run build:index`: Build the documentation index
207- `npm run dev:index`: Build and index in development mode
208- `npm run dev`: Build and start in development mode
209 
210## Troubleshooting
211 
212### Common Issues
213 
2141. **Index not found or failed to load**
215   
216   Run `npm run build:index` to create the index before starting the server.
217 
2182. **API rate limits**
219   
220   When exceeding Google API rate limits, the agent service may return errors. Implement appropriate backoff strategies.
221 
2223. **Model connection issues**
223 
224   Ensure your Google API key is valid and has access to the specified Gemini model.
225 
2264. **Claude Desktop not showing MCP server**
227 
228   - Check your configuration file for syntax errors.
229   - Make sure the path to the server is correct and absolute.
230   - Check Claude Desktop logs for errors.
231   - Restart Claude Desktop after making configuration changes.
232 
233## Contributing
234 
235Contributions are welcome! Please feel free to submit a Pull Request.
236 
237## License
238 
239MIT
240