Borrowing it
Nothing to install: this file belongs to Swih/mistral-mcp. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/Swih/mistral-mcp/main/CLAUDE.mdgit clone --depth 1 https://github.com/Swih/mistral-mcpWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/swih/mistral-mcp/claude-md)<a href="https://agentmods.dev/instructions/swih/mistral-mcp/claude-md"><img src="https://agentmods.dev/badge/instructions/swih/mistral-mcp/claude-md.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.03342 | $0.03342 |
| Opus 5 | $0.01671 | $0.01671 |
| Sonnet 5 | $0.00668 | $0.00668 |
| Haiku 4.5 | $0.00334 | $0.00334 |
Grade A, and why
mistral-mcp CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 8d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 197 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md — mistral-mcp
Instructions de travail pour Claude (Opus 4.7) sur ce repo. Objectif : produire un MCP server Mistral feature-complete et spec-compliant qui peut dormir 6–12 mois sans régression.
1. Identité du projet
- Nom :
mistral-mcp(npm + github.com/Swih/mistral-mcp) - Rôle : serveur MCP stdio + Streamable HTTP wrappant l'API Mistral pour clients MCP (Claude Code, Cursor, Zed, Windsurf, Claude Desktop, ChatGPT Apps).
- Versions en vigueur :
- MCP spec : 2026-07-28, plus le handshake 2025 servi depuis les mêmes
enregistrements. Le second n'est pas de la dette : c'est ce que parlent
quasiment tous les clients diffusés. Toute évolution doit rester servable
aux deux ères —
test/stdio/protocol-eras.test.tsle vérifie. - SDK MCP :
@modelcontextprotocol/server@^2.0.0— toujours via la high-level APIMcpServer+registerTool/Resource/Prompt. Ne jamais descendre au low-levelServer.- stdio et HTTP prennent une factory, jamais une instance :
serveStdio(factory, { legacy: "serve" })etcreateMcpHandler(factory, { legacy: "stateless" }). C'est le SDK qui décide de l'ère. @modelcontextprotocol/sdk@^1.30.0reste en devDependency : c'est le client de l'ère 2025 utilisé pour tester la compat, rien d'autre ne doit l'importer.
- stdio et HTTP prennent une factory, jamais une instance :
- SDK Mistral :
@mistralai/mistralai@^2.6.4(Speakeasy-generated). Retry config obligatoire. - zod :
^4.2.0(plancher imposé par le SDK v2). Attention :.default()porte sur le type de sortie en zod 4 — pour un objet dont les champs ont leurs propres defaults, c'est.prefault({}). - Node :
>=20, TypeScript strict.
- MCP spec : 2026-07-28, plus le handshake 2025 servi depuis les mêmes
enregistrements. Le second n'est pas de la dette : c'est ce que parlent
quasiment tous les clients diffusés. Toute évolution doit rester servable
aux deux ères —
2. Règles dures (ne pas transgresser)
- Zéro claim non-vérifiable. Si une feature n'est pas shippée dans ce repo et testée, elle n'existe pas dans le README, la doc, le CHANGELOG. Pas de "coming soon", pas de roadmap aspirationnelle vendue comme acquise.
- Langues : FR + EN uniquement. Prompts, docs, commentaires, messages d'erreur. Aucune autre langue. Si on trouve de l'espagnol/allemand/etc. dans le code, on strippe.
- Spec-compliance >>> ergonomie. Si la spec MCP dit "include
outputSchema" ou "annotations are REQUIRED for read/write hints", on fait. Toujours. - Chaque tool renvoie
content[]ETstructuredContent.content[]= fallback lisible pour clients pré-2025-06-18.structuredContent= payload JSON strict conforme àoutputSchema. Jamais l'un sans l'autre. - Chaque tool déclare
annotations:title,readOnlyHint,destructiveHint,idempotentHint,openWorldHint. Pour un wrapper API externe,openWorldHint=truequasi toujours. - Erreurs API →
{ content: [text], isError: true }, jamaisthrow. Le LLM appelant doit pouvoir self-correct. - Les inputs zod sont la source de vérité. Jamais de cast
as any, jamais dez.any(). Si un type Mistral n'est pas exporté, on le reconstruit avec zod, on ne "fait confiance". - Le retry config est non-négociable :
backoffstrategy,initialInterval: 500,maxInterval: 5000,exponent: 2,maxElapsedTime: 30000,retryConnectionErrors: true,timeoutMs: 60000. - Pas d'env var autre que
MISTRAL_API_KEYsans justification. L'API key est lue UNE fois danssrc/index.ts. - Pas de dépendance runtime ajoutée sans approbation explicite. Les seules deps runtime acceptées :
@modelcontextprotocol/server,@mistralai/mistralai,zod. Point final. C'est pourquoi l'adaptateurnode:http↔ web-standard est écrit à la main danstransport.tsplutôt que pris chez@modelcontextprotocol/node, qui traîne@hono/node-serverpour quarante lignes.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 8d ago First seen · 197 lines · 3,342 tokens per session scan A 158f17a5e632
mistral-mcp CLAUDE.md is an instructions file published in the GitHub repository Swih/mistral-mcp (15 stars, last pushed 10d ago), licensed MIT. It adds 3,342 tokens to every session, about $0.0167 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
codex AGENTS.md
AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
spec-kit AGENTS.md
AGENTS.md instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.