Content
<div align="center">
# Mistral AI
### Mistral AI's API (La Plateforme): an OpenAI-compatible endpoint for chat and reasoning (Mistral Large/Medium/Small, Magistral), vision (Pixtral), code completion & fill-in-the-middle (Codestral/Devstral), text embeddings, document OCR, and content moderation — plus files, batch, fine-tuning, and agents.
</div>
---
> **Author:** [Simon-Pierre Boucher](https://github.com/simonpierreboucher02) · spbou4@icloud.com
> Part of the **[AgentiLab](https://www.agentilab.ai)** provider knowledge base.
**Categories:** AI Model Providers · Embeddings · Developer Tools
**Tags:** `llm` `mistral` `codestral` `pixtral` `chat` `embeddings` `ocr` `vision` `openai-compatible` `agents`
This repository bundles everything an AI agent (or a developer) needs to use the
**Mistral AI** API: human-readable **documentation**, a ready-to-run **MCP server**, and a
packaged **agent skill**.
## 📊 At a glance
| Asset | Contents |
|-------|----------|
| 📚 **API docs** | 53 markdown pages |
| 🔌 **MCP server** | 7 tools · 31 TypeScript files |
| 🧠 **Agent skill** | 4 recipes · 2 prompts |
| 🗂️ **Total files** | 119 |
| 🔑 **Env vars** | `MISTRAL_API_KEY` |
| 🛡️ **Auth** | Bearer token (Authorization header). Get a key from https://console.mistral.ai. |
| 💳 **Pricing** | Usage-based, billed per token by model (per 1M input/output); embeddings and OCR billed per token/page. A free tier with daily quotas is available; Batch API ~50% cheaper. |
## 📦 What's inside
### 📚 `api-docs/` — Documentation
Structured, example-rich docs covering overview, getting started, the full API
reference, code examples, agent integrations, production patterns and recipes.
Start at [`api-docs/README.md`](api-docs/README.md).
### 🔌 `mcp/` — Model Context Protocol server
A typed MCP server exposing Mistral AI as agent tools:
`mistral_chat` · `mistral_models` · `mistral_embeddings` · `mistral_fim` · `mistral_ocr` · `mistral_moderations` · `mistral_request`
Install, configuration and a tools reference live in
[`mcp/README.md`](mcp/README.md) and [`mcp/docs/`](mcp/docs).
Drop-in client config: [`mcp/examples/claude-desktop-config.json`](mcp/examples/claude-desktop-config.json).
### 🧠 `skill/` — Agent skill
A portable skill (recipes, prompts, reference and evaluation tests) that teaches
an agent how to use Mistral AI effectively. See [`skill/SKILL.md`](skill/SKILL.md).
## 🚀 Quick start (MCP server)
```bash
cd mcp/server
npm install
cp .env.example .env # set MISTRAL_API_KEY
npm run build && npm start
```
Then point any MCP client (e.g. Claude Desktop) at the server using the example config.
## 🔗 Links
- 🌐 Website: https://mistral.ai
- 📖 Official docs: https://docs.mistral.ai
- 🧪 AgentiLab: https://www.agentilab.ai
## 👤 Author
**Simon-Pierre Boucher**
GitHub: [@simonpierreboucher02](https://github.com/simonpierreboucher02) · Email: spbou4@icloud.com
