Skip to main content
Glama
README.md
# Venice MCP Server

[![npm version](https://badge.fury.io/js/venice-mcp.svg)](https://www.npmjs.com/package/venice-mcp)
[![CI](https://github.com/georgeglarson/venice-mcp/actions/workflows/ci.yml/badge.svg)](https://github.com/georgeglarson/venice-mcp/actions/workflows/ci.yml)
[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
[![MCP](https://img.shields.io/badge/MCP-1.12.0-blue)](https://modelcontextprotocol.io)

A [Model Context Protocol](https://modelcontextprotocol.io) (MCP) server for [Venice AI](https://venice.ai). Gives AI assistants like Claude access to Venice's text generation, image creation, text-to-speech, embeddings, and account management APIs.

## What It Does

This MCP server lets Claude (or any MCP-compatible assistant) use Venice AI's features directly:

- **Chat** with Llama, DeepSeek, Qwen, and other open models
- **Generate images** using Stable Diffusion, FLUX, and more
- **Create speech** from text with natural-sounding voices
- **Generate embeddings** for search and RAG applications
- **Manage API keys** and monitor rate limits

## Quick Start

### 1. Get a Venice API Key

Sign up at [venice.ai](https://venice.ai) and create an API key in [Settings → API](https://venice.ai/settings/api).

### 2. Install

```bash
npm install -g venice-mcp
```

### 3. Configure Claude Desktop

Add to your config file:
- **macOS**: `~/Library/Application Support/Claude/claude_desktop_config.json`
- **Windows**: `%APPDATA%\Claude\claude_desktop_config.json`
- **Linux**: `~/.config/Claude/claude_desktop_config.json`

```json
{
  "mcpServers": {
    "venice": {
      "command": "npx",
      "args": ["-y", "venice-mcp"],
      "env": {
        "VENICE_API_KEY": "your-api-key-here"
      }
    }
  }
}
```

Restart Claude Desktop. You'll see Venice tools available in the 🔧 menu.

## Available Tools

### Generation

| Tool | Description |
|------|-------------|
| `venice_chat` | Chat with Venice AI language models |
| `venice_generate_image` | Generate images from text descriptions |
| `venice_upscale_image` | Upscale and enhance images |
| `venice_text_to_speech` | Convert text to spoken audio |
| `venice_create_embeddings` | Generate text embeddings |

### Discovery

| Tool | Description |
|------|-------------|
| `venice_list_models` | List available models (text, image, code) |
| `venice_list_characters` | List character personas |


### Account Management

Requires an admin-level API key:

| Tool | Description |
|------|-------------|
| `venice_list_api_keys` | List all API keys |
| `venice_create_api_key` | Create a new API key |
| `venice_retrieve_api_key` | Get details for a key |
| `venice_delete_api_key` | Delete an API key |
| `venice_get_rate_limits` | View rate limits and usage |
| `venice_get_rate_limit_logs` | View rate limit history |

## Examples

**Ask Claude to generate an image:**
> "Use Venice to create an image of a sunset over mountains"

**Chat with a specific model:**
> "Ask Venice's DeepSeek model to explain quantum computing"

**Check your usage:**
> "Show my Venice API rate limits"

## Supported Models

**Language Models**: Llama 3.3 70B, DeepSeek R1, Qwen 3, Mistral, and more  
**Image Models**: Stable Diffusion 3.5, FLUX, Fluently XL  
**Embeddings**: BGE-M3 (1024 dimensions)  
**TTS**: Kokoro with multiple voice options

Run `venice_list_models` to see all currently available models.

## Requirements

- Node.js 18+
- Venice AI API key

## Development

### Setup

```bash
git clone https://github.com/georgeglarson/venice-mcp.git
cd venice-mcp
npm install
npm run build
```

### Testing

```bash
npm test              # Run tests
npm run test:watch    # Watch mode
npm run test:coverage # Coverage report
```

## Contributing

Contributions are welcome! Please see [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines.

## Links

- [Venice AI](https://venice.ai) - Platform home
- [Venice API Docs](https://docs.venice.ai) - API reference
- [MCP Documentation](https://modelcontextprotocol.io) - Protocol spec
- [GitHub Repository](https://github.com/georgeglarson/venice-mcp)
- [Changelog](CHANGELOG.md)

## License

MIT - see [LICENSE](LICENSE)

TDQS

A3.6/5.0

Scored across 13 tools

Disambiguation5/5

Each tool has a unique purpose (chat, API key CRUD, embeddings, image generation, TTS, etc.) with no overlap, making it easy for an agent to select the correct tool.

Naming Consistency5/5

All tools follow a consistent 'venice_verb_noun' pattern (e.g., venice_create_api_key, venice_list_models), which is predictable and clear.

Tool Count5/5

13 tools covering chat, image, embeddings, TTS, API key management, and monitoring is well-scoped for an AI API server without being excessive.

Completeness4/5

Core functionalities are covered (chat, images, embeddings, TTS, API key CRUD minus update, rate limits). Minor gaps like no streaming or API key update do not hinder basic workflows.

Maintenance

ActivityInactive
ResponsivenessNo issues