Bundle Pluginsource linked

Clawhub Github Publish BkCCGKv0.6.0

MCP server exposing Mistral AI capabilities over MCP: chat, embeddings, FIM, vision, OCR, audio, agents, moderation, classification, files, batch, workflows, sampling, prompts, resources, and Streamable HTTP.

mistral-mcp·runtime mistral-mcp·by @swih
openclaw bundles install clawhub:mistral-mcp
Latest release: v0.6.0Download zip

Capabilities

Bundle format
claude
Host targets
openclaw
Runtime ID
mistral-mcp

Compatibility

Built With Open Claw Version
0.6.0
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Suspicious
medium confidence
Purpose & Capability
The artifacts are coherent with the stated purpose of exposing Mistral AI capabilities over MCP. The surface is broad, including files, batch jobs, workflows, OCR, audio, vision, and agent tools, so users should understand which profile/tools they enable.
Instruction Scope
The curated skills are purpose-aligned, but some automatically read local git diffs when invoked for code review or commit-message generation and then route that content to Mistral tools.
Install Mechanism
There is no top-level ClawHub install spec, while the included plugin/OpenClaw instructions rely on npm/npx, including an npx auto-install with a semver range. This is disclosed but worth pinning in sensitive environments.
Credentials
The Mistral API key requirement is proportional to the integration, but the top-level registry metadata says no required env vars or primary credential even though included skill/plugin artifacts require MISTRAL_API_KEY.
Persistence & Privilege
No hidden local persistence was evident. However, Mistral workflows are described as durable/async and provider-side file, batch, and workflow operations can persist or mutate state in the user's Mistral account.
Scan Findings in Context
[system-prompt-override] expected: The prompt-injection signal appears consistent with included repository/Claude guidance such as CLAUDE.md and skill instructions, not with hidden malicious evaluator manipulation. It should still be treated only as artifact text.
What to consider before installing
Before installing, make sure you are comfortable running the npm package, providing a Mistral API key, and sending selected project data or media to Mistral. In sensitive environments, pin the npm version, keep tool approvals on for delete/batch/workflow actions, and use a revocable key with usage monitoring. ClawScan detected prompt-injection indicators (system-prompt-override), so this skill requires review even though the model response was benign.

Verification

Tier
source linked
Scope
artifact only
Summary
Validated package structure and linked the release to source metadata.
Commit
3dbb79476659
Tag
main
Provenance
No
Scan status
suspicious

Tags

latest
0.6.0

mistral-mcp

MCP server exposing Mistral AI capabilities to any MCP client - Claude Code, Cursor, Zed, Windsurf, Claude Desktop.

Version française : README.fr.md

npm version CI Glama MCP score license MCP spec

Why

Mistral has strong models for French, code, OCR, moderation, audio, and agent-style workflows, but most MCP-enabled IDEs default to Anthropic or OpenAI. mistral-mcp gives those Mistral capabilities a clean MCP surface so you can route the right subtask to the right model without rebuilding your agent loop.

The goal of this repo is not "yet another thin wrapper". It aims to be a robust, maintainable MCP server with explicit schemas, predictable outputs, transport flexibility, and good test coverage.

Current surface (v0.6.0)

Profiles

MISTRAL_MCP_PROFILE controls how many tools are exposed (default: core):

ProfileToolsDescription
core (default)8mistral_chat, mistral_vision, mistral_ocr, codestral_fim, voxtral_transcribe + 3 workflow tools
full25All v0.5 tools + 3 workflow tools. Opt-in for mistral_chat_stream, mistral_embed, mistral_tool_call, etc.
workflows3Workflow tools only — for pipeline orchestration use-cases
MISTRAL_MCP_PROFILE=full node dist/index.js

Tools (25 in full profile — 8 in core)

Core generation:

  • mistral_chat
  • mistral_chat_stream
  • mistral_embed
  • mistral_tool_call
  • codestral_fim

Vision and audio:

  • mistral_vision
  • mistral_ocr
  • voxtral_transcribe
  • voxtral_speak

Agents and classifiers:

  • mistral_agent
  • mistral_moderate
  • mistral_classify

Files and batch:

  • files_upload
  • files_list
  • files_get
  • files_delete
  • files_signed_url
  • batch_create
  • batch_list
  • batch_get
  • batch_cancel

MCP-native utility:

  • mcp_sample - delegates generation to the client model via MCP sampling (full profile)

Workflows (durable execution engine):

  • workflow_execute
  • workflow_status
  • workflow_interact — polymorphic: signal or query against a running execution

Resources (3)

  • mistral://models - accepted aliases and live model catalog
  • mistral://voices - live voice catalog for Voxtral TTS
  • mistral://workflows - live list of deployed Mistral Workflows (use name as workflowIdentifier)

Prompts (6)

French curated prompts:

  • french_invoice_reminder
  • french_meeting_minutes
  • french_email_reply
  • french_commit_message
  • french_legal_summary

English curated prompt:

  • codestral_review

Prompt enum arguments are wrapped with completable(), so MCP clients can call prompt argument completion via completion/complete.

Highlights

  • High-level McpServer API with inputSchema, outputSchema, and annotations on every tool
  • Profile system: MISTRAL_MCP_PROFILE=core|full|workflowscore by default for a lean context footprint
  • Mistral Workflows: workflow_execute / workflow_status / workflow_interact in every profile
  • Dual transport support: stdio by default, Streamable HTTP for remote deployments
  • Structured outputs everywhere: structuredContent plus text fallback
  • OCR annotations: mistral_ocr can request document-level and image/bbox JSON annotations from Mistral Document AI
  • MCP sampling support through mcp_sample (full profile)
  • Prompt completion support for enum-like prompt arguments
  • Resources and prompts registered alongside tools, not bolted on later
  • Retry/backoff and request timeout on the Mistral SDK client

Transport

Stdio

Default mode. This is what Claude Code and most local MCP clients use.

node dist/index.js

Streamable HTTP

Enable with --http or MCP_TRANSPORT=http.

MCP_TRANSPORT=http node dist/index.js

Relevant env vars:

  • MCP_HTTP_HOST - default 127.0.0.1
  • MCP_HTTP_PORT - default 3333
  • MCP_HTTP_PATH - default /mcp
  • MCP_HTTP_TOKEN - optional bearer token
  • MCP_HTTP_ALLOWED_ORIGINS - optional comma-separated allow-list
  • MCP_HTTP_STATELESS=1 - stateless session mode

/healthz is intentionally public and does not touch the MCP server.

Install

Run from npm:

npx mistral-mcp

Or install globally:

npm install -g mistral-mcp
mistral-mcp

Run with Docker:

docker build -t mistral-mcp:dev .
docker run -i --rm -e MISTRAL_API_KEY=your_key_here mistral-mcp:dev

The image uses a multi-stage build and keeps the runtime container to production dependencies plus dist/.

Build from source:

git clone https://github.com/Swih/mistral-mcp.git
cd mistral-mcp
npm install
npm run build

Set your API key:

export MISTRAL_API_KEY=your_key_here

Or use .env at the repo root. Never commit it.

Use in Claude Code

Option A — Claude Code plugin (recommended)

The fastest path: install the bundled Claude Code plugin from the swih-plugins marketplace. It auto-installs the MCP server, prompts for your API key (stored in Claude Code's secrets storage), and ships 5 curated skills.

/plugin marketplace add Swih/mistral-mcp
/plugin install mistral-mcp@swih-plugins

The plugin adds these namespaced skills:

  • /mistral-mcp:mistral-router — picks the right Mistral model + tool for a given task
  • /mistral-mcp:codestral-review — auto-fetches the diff, runs a focused code review
  • /mistral-mcp:french-commit-message — generates a Conventional Commits message in French
  • /mistral-mcp:french-meeting-minutes — audio file or text → structured French meeting minutes
  • /mistral-mcp:french-invoice-reminder — French B2B dunning letter with controlled tone

See claude-plugin/README.md for plugin-specific details.

Option B — Manual MCP server registration

claude mcp add mistral -- node /absolute/path/to/mistral-mcp/dist/index.js

Or via npx, no global install needed:

claude mcp add mistral -- npx -y mistral-mcp@latest

Example prompt:

Use mistral_ocr on this PDF, then run french_meeting_minutes on the extracted text.

Develop

npm run dev
npm run build
npm run lint
npm test
npm run inspector

Test strategy

The suite currently contains 172 tests across 4 layers:

  1. Unit tests for tools, resources, prompts, transport, audio, agents, files, batch, and sampling
  2. Contract tests for tool metadata and MCP-facing guarantees
  3. Live API tests against the real Mistral API when MISTRAL_API_KEY is set
  4. Stdio end-to-end tests against the built server

Without MISTRAL_API_KEY, the local default is 161 passing plus 11 gated live/stdio tests.

Project layout

mistral-mcp/
|-- src/
|   |-- index.ts
|   |-- profile.ts
|   |-- transport.ts
|   |-- tools.ts
|   |-- tools-fn.ts
|   |-- tools-vision.ts
|   |-- tools-audio.ts
|   |-- tools-agents.ts
|   |-- tools-files.ts
|   |-- tools-batch.ts
|   |-- tools-sampling.ts
|   |-- tools-workflows.ts
|   |-- resources.ts
|   `-- prompts.ts
|-- test/
|-- examples/
|-- .github/workflows/ci.yml
|-- package.json
`-- tsconfig.test.json

Status

v0.6.0 — See CHANGELOG.md for the full diff against v0.5.0:

  • profile system (MISTRAL_MCP_PROFILE=core|full|workflows) — core as default for lean tool context
  • 3 new workflow tools: workflow_execute, workflow_status, workflow_interact
  • mistral://workflows resource — live catalog of deployed workflows
  • 172 tests (134 unit + 27 contract + 5 stdio e2e + 6 live API)

Examples

Runnable scripts live in examples/. See examples/README.md.

License

MIT Copyright Dayan Decamp