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

A comprehensive MCP (Model Context Protocol) server providing full access to the Mistral AI API. This server enables AI assistants like Claude to interact with all Mistral AI capabilities including chat completion, embeddings, fine-tuning, agents, OCR, audio transcription, and more.

## Features

### Chat & Completion
- **Chat Completion** - Full chat with function calling, JSON mode, vision support
- **FIM Completion** - Fill-in-the-middle code completion with Codestral
- **Embeddings** - Generate text embeddings for semantic search
- **Agents Completion** - Run completions with pre-configured agents

### Model Management
- List all available models with capabilities
- Get detailed model information
- Delete fine-tuned models
- Update, archive, and unarchive fine-tuned models

### File Management
- Upload files for fine-tuning, batch processing, or OCR
- List, download, and delete files
- Generate signed URLs for file access

### Fine-Tuning
- Create fine-tuning jobs with customizable hyperparameters
- Monitor job status and progress
- Support for dry-run validation
- Cancel and start jobs

### Batch Processing
- Create batch jobs for bulk API calls
- Monitor job progress and status
- Support for chat, embeddings, and FIM endpoints

### Media Processing
- **OCR** - Extract text from documents and images
- **Content Moderation** - Check text and chat for safety issues
- **Classification** - Classify text with fine-tuned classifiers
- **Audio Transcription** - Transcribe audio with Voxtral

### Agents & Conversations (Beta)
- Create and manage agents with custom instructions
- Manage conversations with agents
- Full conversation history support

### Libraries (Beta)
- Create document libraries for RAG
- Upload and manage documents
- Share libraries with users and workspaces

## Installation

```bash
# Clone the repository
git clone https://github.com/firstcastsolutions/mistral-mcp-server.git
cd mistral-mcp-server

# Install dependencies
npm install

# Build the project
npm run build
```

## Configuration

Set the required environment variable:

```bash
export MISTRAL_API_KEY="your-mistral-api-key"
```

## Usage

### With Claude Desktop

Add to your `claude_desktop_config.json`:

```json
{
  "mcpServers": {
    "mistral": {
      "command": "node",
      "args": ["/path/to/mistral-mcp-server/dist/index.js"],
      "env": {
        "MISTRAL_API_KEY": "your-api-key"
      }
    }
  }
}
```

### Standalone

```bash
export MISTRAL_API_KEY="your-api-key"
npm start
```

## Available Tools (40+)

- Chat completion, FIM, embeddings, agents
- Model management (list, get, delete)
- File management (upload, download, delete)
- Fine-tuning jobs
- Batch processing
- OCR, moderation, classification
- Audio transcription
- Agents and conversations (Beta)
- Document libraries (Beta)

## License

MIT - First Cast Solutions LLC

TDQS

C2.9/5.0

Scored across 43 tools

Disambiguation4/5

Tools are mostly distinct, covering different resources (agents, conversations, libraries, files, jobs, models) and operations. Minor potential confusion exists between mistral_get_conversation and mistral_get_conversation_history, and between mistral_download_file and mistral_get_signed_url, but descriptions clarify the differences.

Naming Consistency4/5

Consistent mistral_ prefix and snake_case throughout, with a predictable verb_noun structure for most tools. A few tools (e.g., mistral_embeddings, mistral_ocr, mistral_chat_completion) use noun-based names, slightly breaking the pattern but remaining readable.

Tool Count3/5

With 43 tools, the set is heavy; while the breadth of the Mistral platform justifies many tools, it exceeds typical MCP server scope. Some consolidation might be possible (e.g., file download and signed URL), but each tool maps to a distinct API endpoint.

Completeness4/5

Covers CRUD for agents, conversations, libraries, files, and job management, plus core AI operations like completions, embeddings, OCR, moderation, and transcription. Notable gap: no tools to manage documents within libraries (add/remove/list documents), and some lifecycle operations (e.g., batch job deletion) are missing.

Maintenance

ActivityInactive
ResponsivenessNo issues