# Mistral AI

> Developer documentation for the Mistral AI platform: API references, product guides, and quickstarts for Le Chat, Vibe, Mistral AI Studio, and the Mistral API.

Every link points to the markdown version of a page (same URL with a `.md` extension), served for LLM consumption. Fetch the page URL without the extension for the rendered HTML version.

## Admin

- [API reference](https://docs.mistral.ai/admin/admin-api/api-reference.md): Admin API endpoint reference generated from the OpenAPI specification.
- [Authentication](https://docs.mistral.ai/admin/admin-api/authentication.md): How Admin API keys work, where to create them, and how to authenticate requests.
- [Manage groups and roles](https://docs.mistral.ai/admin/admin-api/manage-groups-roles.md): Create user groups, manage membership and Workspace assignments, and list roles with the Admin API.
- [Manage users](https://docs.mistral.ai/admin/admin-api/manage-users.md): List, create, update, invite, and remove Organization users with the Admin API.
- [Manage Workspaces](https://docs.mistral.ai/admin/admin-api/manage-workspaces.md): List, create, update, and delete Workspaces and manage their members with the Admin API.
- [Overview](https://docs.mistral.ai/admin/admin-api/overview.md): Manage your Organization programmatically: users, groups, Workspaces, API keys, limits, usage, and audit logs.
- [Usage metrics](https://docs.mistral.ai/admin/admin-api/usage-metrics.md): Retrieve billing usage, spending and rate limits, le Chat activity, and Vibe Code analytics with the Admin API.
- [User provisioning](https://docs.mistral.ai/admin/admin-api/user-provisioning.md): Bulk-create or invite users, organize them into a user group, and grant them access to a Workspace with the Admin API.
- [Billing](https://docs.mistral.ai/admin/billing-usage/billing.md): Use Billing to manage how your Organization pays for Mistral services.
- [Invoices](https://docs.mistral.ai/admin/billing-usage/invoices.md): Find and download your invoices.
- [Subscriptions](https://docs.mistral.ai/admin/billing-usage/subscriptions.md): Use Subscription to review the Mistral plan, seats, included monthly usage, and pay-as-you-go settings active on your Organization.
- [Usage and limits](https://docs.mistral.ai/admin/billing-usage/usage-limits.md): Monitor Organization usage, understand limits, and decide where to adjust caps.
- [API keys](https://docs.mistral.ai/admin/identity-access/api-keys.md): API keys authenticate requests to the Mistral API and other Mistral tools.
- [Connectors](https://docs.mistral.ai/admin/identity-access/connectors.md): Manage organization app connections and Connector tool access from Admin.
- [Groups](https://docs.mistral.ai/admin/identity-access/groups.md): Group users to manage Workspace access and roles at scale.
- [Roles and permissions (RBAC)](https://docs.mistral.ai/admin/identity-access/roles-permissions.md): Predefined Organization and Workspace roles, what each one grants, and how to assign them.
- [Service accounts](https://docs.mistral.ai/admin/identity-access/service-accounts.md): Give a workload its own non-human identity in a workspace, with its own roles and its own credentials.
- [User management](https://docs.mistral.ai/admin/identity-access/user-management.md): Manage members, invitations, Organization roles, seats, Workspaces, groups, and user API keys.
- [Workload identity federation](https://docs.mistral.ai/admin/identity-access/workload-identity.md): Let a workload authenticate as a service account with short-lived tokens signed by your own cluster, instead of a stored secret.
- [Overview](https://docs.mistral.ai/admin/monitor-comply/audit-logs/overview.md): Audit logs provide a chronological record of actions performed within your Organization across Studio and Vibe.
- [Audit logs reference](https://docs.mistral.ai/admin/monitor-comply/audit-logs/reference.md): Audit log response fields, filters, actor types, target types, and event types.
- [Zero data retention](https://docs.mistral.ai/admin/monitor-comply/zero-data-retention.md): Learn what zero data retention covers, which API endpoints are eligible, and how to request it for your Organization.
- [Admin overview](https://docs.mistral.ai/admin/overview.md): The Admin Panel is the control surface for your Organization.
- [Admin](https://docs.mistral.ai/admin.md): Manage security, access controls, users, billing, and organizational settings for your Mistral Workspace.
- [Create your Organization](https://docs.mistral.ai/admin/set-up-organization/create-organization.md): Create your Mistral Organization and add your first members.
- [Enterprise Accounts](https://docs.mistral.ai/admin/set-up-organization/enterprise-accounts.md): Understand how Enterprise Accounts fit above Organizations.
- [Email domain authentication](https://docs.mistral.ai/admin/set-up-organization/sign-in-method/email-domain-authentication.md): Email domain authentication is available on Team plans and above.
- [SAML SSO](https://docs.mistral.ai/admin/set-up-organization/sign-in-method/saml-sso.md): SAML single sign-on (SSO) lets members of your Organization sign in with your corporate identity provider (IdP).
- [Verify your domain](https://docs.mistral.ai/admin/set-up-organization/verify-domain.md): Verify domain ownership with a DNS record.
- [Workspace usage and limits](https://docs.mistral.ai/admin/workspaces/usage-limits.md): Monitor Workspace consumption and set spending caps and rate limits.
- [Workspaces in Studio](https://docs.mistral.ai/admin/workspaces/workspaces-in-studio.md): Understand what Workspaces change for Studio users and what admins can configure.
- [Workspaces in Vibe](https://docs.mistral.ai/admin/workspaces/workspaces-in-vibe.md): Understand what Workspaces change for Vibe users and what admins can configure.
- [Your first Workspace](https://docs.mistral.ai/admin/workspaces/your-first-workspace.md): Understand what a Workspace is and create one inside your Organization.

## Getting Started

- [SDK Clients](https://docs.mistral.ai/getting-started/clients.md): ExplorerTabItem,
- [Platform overview](https://docs.mistral.ai/getting-started/platform-overview.md): Three products to use Mistral AI: Vibe for daily work and coding, Studio for the API and Playground, Admin for organization-level control.
- [Configure SSO and domain verification](https://docs.mistral.ai/getting-started/quickstarts/admin/configure-sso.md): Verify your domain and set up SAML SSO so team members sign in with your corporate identity provider. ~20 minutes.
- [Create your organization](https://docs.mistral.ai/getting-started/quickstarts/admin/create-organization.md): Create your organization, configure security policies, and invite your first team member in ~15 minutes.
- [Manage workspaces and API keys](https://docs.mistral.ai/getting-started/quickstarts/admin/manage-workspaces.md): Create isolated workspaces, generate API keys, and set usage limits for your team. ~10 minutes.
- [Build a workflow](https://docs.mistral.ai/getting-started/quickstarts/developer/build-a-workflow.md): Scaffold a durable AI workflow, run a worker, and trigger your first execution in ~15 minutes.
- [Build an agent with tools](https://docs.mistral.ai/getting-started/quickstarts/developer/build-an-agent.md): Create an AI agent that calls external functions and returns real-world data in ~10 minutes.
- [Send your first API request](https://docs.mistral.ai/getting-started/quickstarts/developer/first-api-request.md): Send your first request to the Mistral API and get a model response in ~5 minutes.
- [Set up RAG with document search](https://docs.mistral.ai/getting-started/quickstarts/developer/rag-document-search.md): Upload documents to a knowledge Library and query them with retrieval-augmented generation in ~10 minutes.
- [Activate Studio and generate an API key](https://docs.mistral.ai/getting-started/quickstarts/studio/activate-and-generate-api-key.md): Generate your first API key in Studio in ~5 minutes.
- [Create a reusable Prompt](https://docs.mistral.ai/getting-started/quickstarts/studio/create-reusable-prompt.md): Save a prompt template in Studio and choose whether to keep it private or share it with your workspace in ~5 minutes.
- [Create a Skill in Studio](https://docs.mistral.ai/getting-started/quickstarts/studio/create-skill.md): Package reusable instructions and files into a Skill from Studio in ~10 minutes.
- [Test a model in the API playground](https://docs.mistral.ai/getting-started/quickstarts/studio/test-model-playground.md): Send prompts, adjust parameters, and compare model outputs in the Studio playground in ~5 minutes.
- [Install the Vibe CLI and send your first prompt](https://docs.mistral.ai/getting-started/quickstarts/vibe-code/install-cli.md): Install the Vibe CLI, configure your API key, and send your first prompt from the terminal. ~5 minutes.
- [Scaffold a project with Vibe Code](https://docs.mistral.ai/getting-started/quickstarts/vibe-code/scaffold-a-project.md): Use Vibe Code to scaffold a new project from a natural-language description, reviewing every change before it's applied. ~10 minutes.
- [Analyze a dataset](https://docs.mistral.ai/getting-started/quickstarts/vibe-work/analyze-data.md): Upload a spreadsheet to Vibe Work and ask questions in plain language.
- [Create your first Skill](https://docs.mistral.ai/getting-started/quickstarts/vibe-work/create-first-skill.md): Package a reusable method into a Skill so Vibe Work applies the same procedure every time the task matches. ~10 minutes.
- [Run your first Vibe Work task](https://docs.mistral.ai/getting-started/quickstarts/vibe-work/first-task.md): Open Vibe Work, run a small multi-step task end-to-end, and review the result. ~5 minutes.

## Inference

- [Labs](https://docs.mistral.ai/inference/labs.md): Labs are experimental and fast moving models available for a limited amount of time.
- [Model lifecycle policy](https://docs.mistral.ai/inference/model-lifecycle.md): Each model on the Mistral Serverless API moves through a defined lifecycle, from initial release to retirement.
- [Inference](https://docs.mistral.ai/inference.md): Explore model serving, model choice, request behavior, and inference-related API concepts.
- [Priority Tier](https://docs.mistral.ai/inference/priority-tier.md): Priority Tier gives eligible API requests priority queueing for workloads that need more predictable access to shared infrastructure.
- [Prompting](https://docs.mistral.ai/inference/prompting.md): When you first start using Mistral models, your initial interaction will revolve around prompts.
- [Regional inference](https://docs.mistral.ai/inference/regional-inference.md): Mistral offers regional inference as an optional service through dedicated API endpoints.
- [Sampling](https://docs.mistral.ai/inference/sampling.md): Here, we will discuss the sampling settings that influence the output of Large Language Models (LLMs).
- [API pricing](https://docs.mistral.ai/inference/pricing.md): Pricing for Mistral models through the API, organized by family.
- [Model Selection Guide](https://docs.mistral.ai/inference/model-selection-guide.md): Compare current Mistral models on speed, performance, modalities, context size, pricing and license.

## Models

- [Models Overview](https://docs.mistral.ai/models.md): Every Mistral model with its API name, status and context length, including deprecated and retired models.
- [Z.ai GLM 5.3](https://docs.mistral.ai/models/zai-glm-5-3.md): A third-party open weight text model from Z.ai with a 1M-token context window.
- [Z.ai GLM 5.2](https://docs.mistral.ai/models/zai-glm-5-2.md): A third-party open weight text model from Z.ai with a 1M-token context window.
- [Shieldstral 1.0](https://docs.mistral.ai/models/shieldstral-1-0.md): Compact multimodal moderation model for text and image safety classification.
- [Timestral 1.0 209M](https://docs.mistral.ai/models/timestral-1-0-209m.md): A time series foundation model for zero-shot forecasting with covariate support and probabilistic outputs.
- [Timestral 1.0 123M](https://docs.mistral.ai/models/timestral-1-0-123m.md): A time series foundation model for zero-shot forecasting with covariate support and probabilistic outputs.
- [Timestral 1.0 28M](https://docs.mistral.ai/models/timestral-1-0-28m.md): A time series foundation model for zero-shot forecasting with covariate support and probabilistic outputs.
- [Timestral 1.0 5M](https://docs.mistral.ai/models/timestral-1-0-5m.md): A time series foundation model for zero-shot forecasting with covariate support and probabilistic outputs.
- [Mistral Medium 3.5](https://docs.mistral.ai/models/mistral-medium-3-5-26-04.md): Our frontier-class multimodal model optimized for agentic and coding use cases.
- [Voxtral TTS](https://docs.mistral.ai/models/voxtral-tts-26-03.md): State-of-the-art TTS with zero-shot voice cloning and multilingual support.
- [Voxtral Mini Transcribe 3](https://docs.mistral.ai/models/voxtral-mini-transcribe-3-0.md): Strong transcription and diarization at a competitive price point.
- [Voxtral Small Transcribe 3](https://docs.mistral.ai/models/voxtral-small-transcribe-3-0.md): Flagship transcription and diarization model with SOTA accuracy.
- [Voxtral Mini Transcribe Realtime 3](https://docs.mistral.ai/models/voxtral-mini-transcribe-realtime-3-0.md): Realtime transcription model for agentic use-cases with configurable latency and turn detection.
- [Voxtral TTS 3](https://docs.mistral.ai/models/voxtral-mini-tts-3-0.md): State-of-the-art TTS with zero-shot and professional voice cloning, 9 languages, and streaming.
- [Mistral Small 4](https://docs.mistral.ai/models/mistral-small-4-0-26-03.md): Hybrid model unifying instruct, reasoning, and coding in a single efficient model.
- [Leanstral 1.5](https://docs.mistral.ai/models/leanstral-1-5.md): Updated code agent for Lean 4 formal proof engineering and automated theorem proving.
- [Leanstral](https://docs.mistral.ai/models/leanstral-26-03.md): Open-source code agent for Lean 4 formal proof engineering.
- [Mistral Large 4](https://docs.mistral.ai/models/mistral-large-4-0.md): A state-of-the-art, open-weight, general-purpose multimodal model.
- [Mistral Large 3](https://docs.mistral.ai/models/mistral-large-3-25-12.md): A state-of-the-art, open-weight, general-purpose multimodal model.
- [Mistral Medium 3.1](https://docs.mistral.ai/models/mistral-medium-3-1-25-08.md): Our frontier-class multimodal model released August 2025.
- [Mistral Small 3.2](https://docs.mistral.ai/models/mistral-small-3-2-25-06.md): An update to our previous small model, released June 2025.
- [OCR 4.1](https://docs.mistral.ai/models/ocr-4-1.md): Our latest OCR service with paragraph-level bounding boxes, structural block labels, and block-level confidence scores.
- [OCR 4.0](https://docs.mistral.ai/models/ocr-4-0.md): Our latest OCR service with paragraph-level bounding boxes and structural block labels.
- [OCR 3](https://docs.mistral.ai/models/ocr-3-25-12.md): Our OCR service powering our Document AI stack. OCR 4 is available as the newer model. OCR 3 remains available for existing integrations and production workloads.
- [Voxtral Mini Transcribe](https://docs.mistral.ai/models/voxtral-mini-transcribe-25-07.md): An efficient audio input model, pre-trained and optimized for transcription purposes.
- [Voxtral Mini Transcribe 2](https://docs.mistral.ai/models/voxtral-mini-transcribe-26-02.md): An efficient audio input model, pre-trained and optimized for transcription purposes.
- [Voxtral Mini Transcribe Realtime](https://docs.mistral.ai/models/voxtral-mini-transcribe-realtime-26-02.md): An efficient audio input model, pre-trained and optimized for live transcription purposes.
- [Codestral](https://docs.mistral.ai/models/codestral-25-08.md): Our cutting-edge language model for code completion released end of July 2025.
- [Devstral 2](https://docs.mistral.ai/models/devstral-2-25-12.md): Our frontier code agents model for solving software engineering tasks.
- [Magistral Medium 1.1](https://docs.mistral.ai/models/magistral-medium-1-1-25-07.md): Our frontier-class reasoning model released July 2025.
- [Codestral Embed](https://docs.mistral.ai/models/codestral-embed-25-05.md): Our state-of-the-art semantic for extracting representation of code extracts
- [Mistral Small Creative](https://docs.mistral.ai/models/mistral-small-creative-25-12.md): A specialized model designed for creative writing and dynamic character interaction.
- [Devstral Small 2](https://docs.mistral.ai/models/devstral-small-2-25-12.md): Model that excels at using tools to explore codebases, editing multiple files and power software engineering agents.
- [Ministral 3 14B](https://docs.mistral.ai/models/ministral-3-14b-25-12.md): A powerful model offering best-in-class text and vision capabilities. 
- [Ministral 3 8B](https://docs.mistral.ai/models/ministral-3-8b-25-12.md): A powerful and efficient model offering best-in-class text and vision capabilities. 
- [Ministral 3 3B](https://docs.mistral.ai/models/ministral-3-3b-25-12.md): A tiny and efficient model offering best-in-class text and vision capabilities. 
- [Magistral Medium 1.2](https://docs.mistral.ai/models/magistral-medium-1-2-25-09.md): Our frontier-class multimodal reasoning model.
- [Magistral Small 1.2](https://docs.mistral.ai/models/magistral-small-1-2-25-09.md): Our small multimodal reasoning model.
- [Magistral Small 1.1](https://docs.mistral.ai/models/magistral-small-1-1-25-07.md): Our small reasoning model released July 2025.
- [Voxtral Mini](https://docs.mistral.ai/models/voxtral-mini-25-07.md): A mini version of our first audio input model.
- [Voxtral Small](https://docs.mistral.ai/models/voxtral-small-25-07.md): Our first model with audio input capabilities for instruct use cases.
- [Devstral Medium 1.0](https://docs.mistral.ai/models/devstral-medium-1-0-25-07.md): An enterprise grade text model, that excels at SWE use cases.
- [Devstral Small 1.1](https://docs.mistral.ai/models/devstral-small-1-1-25-07.md): An update to our open source model that excels at SWE use cases.
- [Magistral Medium 1.0](https://docs.mistral.ai/models/magistral-medium-1-0-25-06.md): Our first frontier-class reasoning model released June 2025
- [Magistral Small 1.0](https://docs.mistral.ai/models/magistral-small-1-0-25-06.md): Our first small reasoning model released June 2025.
- [OCR 2](https://docs.mistral.ai/models/ocr-2-25-05.md): Our OCR service powering our Document AI stack.
- [Devstral Small 1.0](https://docs.mistral.ai/models/devstral-small-1-0-25-05.md): A 24B text model, open source model that excels at SWE use cases.
- [Mistral Medium 3](https://docs.mistral.ai/models/mistral-medium-3-25-05.md): Our frontier-class multimodal model released May 2025.
- [Mistral Small 3.1](https://docs.mistral.ai/models/mistral-small-3-1-25-03.md): A new leader in the small models category with image understanding capabilities.
- [OCR](https://docs.mistral.ai/models/ocr-25-03.md): Our first OCR optimized model.
- [Mistral Saba](https://docs.mistral.ai/models/mistral-saba-25-02.md): A finetuned Small model trained on meticulously curated datasets.
- [Mistral Small 3.0](https://docs.mistral.ai/models/mistral-small-3-0-25-01.md): A small efficient and powerfull 24B open model for personal or commercial use.
- [Codestral](https://docs.mistral.ai/models/codestral-25-01.md): Our cutting-edge language model for coding.
- [Mistral Large 2.1](https://docs.mistral.ai/models/mistral-large-2-1-24-11.md): Our top-tier large model for high-complexity tasks with the lastest version released November 2024.
- [Pixtral Large](https://docs.mistral.ai/models/pixtral-large-24-11.md): Our first frontier-class multimodal model released November 2024.
- [Mistral Moderation 2](https://docs.mistral.ai/models/mistral-moderation-26-03.md): Our latest moderation model with 128k context window and jailbreaking detection.
- [Mistral Moderation](https://docs.mistral.ai/models/mistral-moderation-24-11.md): Our moderation service that enables our users to detect harmful text content.
- [Ministral 3B](https://docs.mistral.ai/models/ministral-3b-24-1.md): World’s best edge model.
- [Ministral 8B](https://docs.mistral.ai/models/ministral-8b-24-1.md): Powerful edge model with extremely high performance/price ratio.
- [Mistral Small 2.0](https://docs.mistral.ai/models/mistral-small-2-0-24-09.md): An update to our efficient small model with open weights under an MRL license.
- [Pixtral 12B](https://docs.mistral.ai/models/pixtral-12b-24-09.md): A 12B model with image understanding capabilities in addition to text.
- [Mistral Large 2.0](https://docs.mistral.ai/models/mistral-large-2-0-24-07.md): A powerfull model with open weights under a MRL license.
- [Mistral Nemo 12B](https://docs.mistral.ai/models/mistral-nemo-12b-24-07.md): Our best multilingual open source model released July 2024.
- [Codestral Mamba 7B](https://docs.mistral.ai/models/codestral-mamba-7b-0-1.md): A coding dedicated model using a mamba 2 architecture for research purposes.
- [Mathstral 7B](https://docs.mistral.ai/models/mathstral-7b-0-1.md): A math dedicated model for research purposes.
- [Codestral](https://docs.mistral.ai/models/codestral-24-05.md): Our first cutting-edge language model for coding and FIM.
- [Mistral 7B](https://docs.mistral.ai/models/mistral-7b-0-3.md): A minor update to our first open model.
- [Mixtral 8x22B](https://docs.mistral.ai/models/mixtral-8x22b-0-1-0-3.md): A large open MoE model competing with much larger dense models.
- [Mistral Small 1.0](https://docs.mistral.ai/models/mistral-small-1-0-24-02.md): An enterprise grade small model for efficiency.
- [Mistral Large 1.0](https://docs.mistral.ai/models/mistral-large-1-0-24-02.md): Our first and larger SOTA model for enterprise use cases.
- [Mistral Next](https://docs.mistral.ai/models/mistral-next.md): A prototype model, designed to be brief and concise.
- [Mistral Embed](https://docs.mistral.ai/models/mistral-embed-23-12.md): Our state-of-the-art semantic for extracting representation of code extracts
- [Mistral Medium 1.0](https://docs.mistral.ai/models/mistral-medium-1-0-23-12.md): Our first SOTA enterprise grade model.
- [Mixtral 8x7B](https://docs.mistral.ai/models/mixtral-8x7b-0-1.md): Our first open MoE model, SOTA competing against models up to 2 to 3x larger.
- [Mistral 7B](https://docs.mistral.ai/models/mistral-7b-0-2.md): An update to our first open model with better context.
- [Mistral 7B](https://docs.mistral.ai/models/mistral-7b-0-1.md): Our very first open model, SOTA competing against models up to 2 to 3x larger.

## Resources

- [Fine-tuning API (legacy)](https://docs.mistral.ai/resources/deprecated/customization.md): This feature is deprecated and is no longer actively supported.
- [Classifier Factory](https://docs.mistral.ai/resources/deprecated/finetuning/classifier_factory.md): This feature is deprecated and is no longer actively supported.
- [Fine-tuning](https://docs.mistral.ai/resources/deprecated/finetuning.md): This feature is deprecated and is no longer actively supported.
- [Text & Vision Fine-tuning](https://docs.mistral.ai/resources/deprecated/finetuning/text_vision_finetuning.md): This feature is deprecated and is no longer actively supported.
- [Mistral Moderation 2411](https://docs.mistral.ai/resources/deprecated/guardrailing/mistral_moderation_2411.md): Deprecated: mistral-moderation-2411 is deprecated.
- [Safe Prompt](https://docs.mistral.ai/resources/deprecated/guardrailing/safe_prompt.md): safe_prompt is deprecated.
- [Native reasoning (deprecated)](https://docs.mistral.ai/resources/deprecated/native-reasoning.md): Native reasoning models (magistral-small-latest, magistral-medium-latest) are deprecated.
- [Error glossary](https://docs.mistral.ai/resources/error-glossary.md): This page lists the HTTP status codes returned by the Mistral API, their meanings, and how to resolve them.
- [Glossary](https://docs.mistral.ai/resources/glossary.md): This glossary defines the key terms used throughout our documentation: models, API concepts, deployment options, and platform features.
- [Known limitations](https://docs.mistral.ai/resources/known-limitations.md): This page documents current limitations of the Mistral platform.
- [Supported languages](https://docs.mistral.ai/resources/languages.md): Mistral's language models are trained on multilingual data and support a wide range of languages.
- [MCP](https://docs.mistral.ai/resources/mcp.md): Connect any MCP client to Mistral Studio with your API key.
- [Migration guides](https://docs.mistral.ai/resources/migration-guides.md): Migrate to Mistral from OpenAI or self-hosted Llama with minimal code changes.
- [Observability integrations](https://docs.mistral.ai/resources/observability-integrations.md): Connect Mistral traces, evaluations, and feedback to third-party observability tools.
- [Resources](https://docs.mistral.ai/resources.md): Find SDKs, cookbooks, release notes, migration guides, glossary entries, and support references for building with Mistral.
- [Release notes](https://docs.mistral.ai/resources/release-notes.md): Follow shipped product updates across Vibe, Studio, Admin, models, and the API.
- [SDKs](https://docs.mistral.ai/resources/sdks.md): ExplorerTabItem,
- [Mistral AI Studio API files input filtering and access restriction](https://docs.mistral.ai/resources/security-advisories/MAI-2026-001.md): Security advisory for API files input filtering and access restriction in Mistral AI Studio.
- [TanStack supply chain attack affecting Mistral SDK packages](https://docs.mistral.ai/resources/security-advisories/MAI-2026-002.md): Security advisory for the TanStack supply chain attack affecting Mistral SDK packages.
- [Mistral Vibe shell permission vulnerabilities](https://docs.mistral.ai/resources/security-advisories/MAI-2026-003.md): Security advisory for shell permission vulnerabilities affecting Mistral Vibe.
- [Security advisories](https://docs.mistral.ai/resources/security-advisories.md): Security advisories and remediation guidance for incidents affecting Mistral SDKs, packages, or developer tooling.
- [Changelog](https://docs.mistral.ai/resources/changelogs.md): Dated changes to Mistral models, the API and products, newest first.

## Studio

- [Code Interpreter](https://docs.mistral.ai/studio/agents/agent-tools/code_interpreter.md): code_interpreter works with the Conversations API (/v1/conversations) and the Agents API.
- [Function Calling](https://docs.mistral.ai/studio/agents/agent-tools/function-calling.md): The core of an agent relies on its tool usage capabilities, enabling it to use and call tools and workflows depending on the task it must accomplish.
- [Image Generation](https://docs.mistral.ai/studio/agents/agent-tools/image_generation.md): Enabling this tool allows models to create images at any given moment.
- [Agents Tools Overview](https://docs.mistral.ai/studio/agents/agent-tools.md): Agents can use tools to interact with the external world, these can be APIs, databases, or other services, increasing the capabilities of your agent and extending its functionality beyond its own knowledge base and fi...
- [Websearch](https://docs.mistral.ai/studio/agents/agent-tools/websearch.md): Websearch is the capability to browse the web in search of information, this tool does not only fix the limitations of models of not being up to date due to their training data, but also allows them to actually retrie...
- [Agents & Conversations](https://docs.mistral.ai/studio/agents/agents-api.md): Agents is a feature that allows developers to create predefined models with their own system prompts and tools.
- [Handoffs](https://docs.mistral.ai/studio/agents/handoffs.md): When creating and using Agents, often with access to specific tools, there are moments where it is desired to call other Agents mid-action.
- [Agents Introduction](https://docs.mistral.ai/studio/agents/introduction.md): AI agents are autonomous systems powered by large language models (LLMs) that, given high-level instructions, can plan, use tools, carry out processing steps, and take actions to achieve specific goals.
- [Audio in chat completions](https://docs.mistral.ai/studio/audio/audio-in-chat.md): Summarize, question, translate, and extract structured data from a recording by sending the audio in a chat completions request.
- [Custom vocabulary](https://docs.mistral.ai/studio/audio/concepts/custom-vocabulary.md): Why names, drugs, and product terms come out wrong in transcription and in generated speech, and the hint that fixes each direction: context biasing for spelling, custom pronunciation for sound.
- [Latency](https://docs.mistral.ai/studio/audio/concepts/latency.md): What latency means in each Mistral audio mode, the delays it adds up, and how to measure it and weigh speed against quality.
- [Offline and realtime audio](https://docs.mistral.ai/studio/audio/concepts/offline-and-realtime.md): The two ways Mistral processes audio: recorded audio in a single request, or live audio in a streaming session, and what changes between them.
- [Speakers and diarization](https://docs.mistral.ai/studio/audio/concepts/speakers-and-diarization.md): What diarization tells you about who spoke when, how reference clips turn anonymous speaker labels into known speakers, and what overlapping speech means for your interface.
- [Turns and turn-taking](https://docs.mistral.ai/studio/audio/concepts/turns.md): How Voxtral decides when someone has finished speaking, why it matters for live transcription and voice agents, and the trade-off between cutting people off and answering late.
- [Voices](https://docs.mistral.ai/studio/audio/concepts/voices.md): What a voice is in Mistral speech generation, the kinds of voices you can use, and the difference between choosing a voice and directing how it speaks.
- [Audio](https://docs.mistral.ai/studio/audio/deprecated/overview.md): Transcribe speech, generate expressive voices, and build real-time voice agents with Mistral's Voxtral audio models.
- [Offline](https://docs.mistral.ai/studio/audio/deprecated/speech_to_text/offline_transcription.md): Before You Start
- [Speech to Text](https://docs.mistral.ai/studio/audio/deprecated/speech_to_text.md): This page covers Mistral's Speech Transcription capabilities, including Offline and Realtime transcription.
- [Client authentication](https://docs.mistral.ai/studio/audio/deprecated/speech_to_text/realtime_transcription/client_auth.md): Browser clients cannot store long-lived API keys safely and cannot set Authorization headers on WebSocket connections.
- [Realtime](https://docs.mistral.ai/studio/audio/deprecated/speech_to_text/realtime_transcription.md): Realtime transcription allows you to transcribe audio as it is being spoken or recorded.
- [Text to Speech](https://docs.mistral.ai/studio/audio/deprecated/text_to_speech.md): Voxtral TTS is Mistral's text-to-speech model with zero-shot voice cloning.
- [Speech Generation](https://docs.mistral.ai/studio/audio/deprecated/text_to_speech/speech.md): Generate speech from text using a saved voice (voice_id) or a one-off reference audio clip (ref_audio).
- [Voices](https://docs.mistral.ai/studio/audio/deprecated/text_to_speech/voices.md): Save audio samples as reusable voices.
- [Migrate from Voxtral 2](https://docs.mistral.ai/studio/audio/migrate-from-voxtral-2.md): Move from the Voxtral 2 audio models to Voxtral 3: model names, breaking changes, and a migration checklist.
- [Audio](https://docs.mistral.ai/studio/audio/overview.md): Transcribe speech, generate expressive voices, and build real-time voice agents with Mistral's Voxtral audio models.
- [Audio formats](https://docs.mistral.ai/studio/audio/specs-and-limits/audio-formats.md): Accepted input formats and available output formats across transcription, speech generation, audio in chat, and voice agents.
- [Languages](https://docs.mistral.ai/studio/audio/specs-and-limits/languages.md): How each Voxtral 3 model handles languages for transcription, speech generation, and translation, and where to find its supported languages.
- [Limits](https://docs.mistral.ai/studio/audio/specs-and-limits/limits.md): Size, duration, and value limits for audio requests and sessions, and what happens when you go over them.
- [Bias context and vocabulary](https://docs.mistral.ai/studio/audio/speech-to-text/bias-context-and-vocabulary.md): Fix jargon, product names, and domain terms with keyphrases and free-form context.
- [Diarize speakers](https://docs.mistral.ai/studio/audio/speech-to-text/diarize-speakers.md): Label who speaks in a recorded file or a live session, read the speaker labels, and name known speakers with reference clips.
- [Speech to Text](https://docs.mistral.ai/studio/audio/speech-to-text/overview.md): Choose the right transcription model, mode, and configuration.
- [Quickstart: transcribe an audio file](https://docs.mistral.ai/studio/audio/speech-to-text/quickstarts/transcribe-a-file.md): Turn an audio file into accurate text with a single API call.
- [Quickstart: transcribe speech in realtime](https://docs.mistral.ai/studio/audio/speech-to-text/quickstarts/transcribe-in-realtime.md): Stream live audio and get a transcript with low latency.
- [File transcription](https://docs.mistral.ai/studio/audio/speech-to-text/transcribe-files.md): Transcribe recorded audio: send a file, a URL, or an uploaded file, pin the language, get segment timestamps, and stream the response.
- [Realtime transcription](https://docs.mistral.ai/studio/audio/speech-to-text/transcribe-in-realtime.md): Transcribe live audio over a WebSocket session: the session lifecycle, its settings, the events to read, and how to trade speed for accuracy.
- [Speech to Text troubleshooting](https://docs.mistral.ai/studio/audio/speech-to-text/troubleshooting.md): The errors that transcription requests and realtime sessions return, the usual causes of a poor or missing transcript, and the setting that fixes each one.
- [Turn detection](https://docs.mistral.ai/studio/audio/speech-to-text/turn-detection.md): Enable end-of-turn detection in realtime transcription, read the turn events and their timing against the text, and tune the thresholds.
- [Custom pronunciation](https://docs.mistral.ai/studio/audio/text-to-speech/custom-pronunciation.md): Force the pronunciation of brand names, product names, and jargon with custom phonemes.
- [Emotion and pace tags](https://docs.mistral.ai/studio/audio/text-to-speech/emotion-tags.md): Change the emotion and the speaking rate of part of the text with inline tags in the speech input.
- [Speech generation](https://docs.mistral.ai/studio/audio/text-to-speech/generate-speech.md): Turn text into audio with control over voice, format, and expressiveness.
- [Text to Speech](https://docs.mistral.ai/studio/audio/text-to-speech/overview.md): Choose the speech generation mode, model, and features that fit your application.
- [Quickstart: generate speech](https://docs.mistral.ai/studio/audio/text-to-speech/quickstarts/generate-speech.md): Convert text into natural speech and play it back with one API call.
- [Stream speech](https://docs.mistral.ai/studio/audio/text-to-speech/stream-speech.md): Stream the audio of a speech request as it's generated, play it before the whole file is ready, and measure the time to first audio.
- [Style instructions](https://docs.mistral.ai/studio/audio/text-to-speech/style-instructions.md): Steer the delivery of the whole speech output with a short free-form instruction.
- [Text to Speech troubleshooting](https://docs.mistral.ai/studio/audio/text-to-speech/troubleshooting.md): The errors that speech requests return, the usual causes of speech that sounds wrong, flat, cut short, or slow to start, and the setting that fixes each one.
- [Voice cloning](https://docs.mistral.ai/studio/audio/text-to-speech/voice-cloning.md): Clone a voice from a short recording, for one request or saved in your workspace.
- [Voices](https://docs.mistral.ai/studio/audio/text-to-speech/voices.md): Browse the preset voices and manage the voices saved in your workspace through the API.
- [Create and manage agents](https://docs.mistral.ai/studio/audio/voice-agents/agents.md): Save and version agent configurations, then start sessions from them.
- [Audio interface](https://docs.mistral.ai/studio/audio/voice-agents/audio-interface.md): How a voice agent captures and plays audio in Python and Node.js, how to tune the default audio interface, write your own, or send and receive raw audio.
- [Browser (TypeScript) guide](https://docs.mistral.ai/studio/audio/voice-agents/browser.md): Run voice agents in the browser with a server-assisted session.
- [Agent configuration](https://docs.mistral.ai/studio/audio/voice-agents/configuration.md): The Mistral client and the audio.sessions.init() options — agent configuration vs session runtime.
- [Custom audio interfaces](https://docs.mistral.ai/studio/audio/voice-agents/custom-audio.md): Take control of audio I/O: subclass the interface, or send and receive raw audio manually.
- [Voice agents](https://docs.mistral.ai/studio/audio/voice-agents/overview.md): Real-time bidirectional voice conversations through the Voice Agents SDK: microphone capture, STT, LLM, TTS, and playback handled by the SDK.
- [Quickstart: build a voice agent](https://docs.mistral.ai/studio/audio/voice-agents/quickstarts/build-a-voice-agent.md): Run a voice conversation with an agent from your terminal in a few steps.
- [Manage agent sessions](https://docs.mistral.ai/studio/audio/voice-agents/sessions.md): Session lifecycle, the event model, and error handling for production voice agents.
- [Telephony (SIP)](https://docs.mistral.ai/studio/audio/voice-agents/telephony.md): Answer phone calls with your voice agent over SIP.
- [Tools](https://docs.mistral.ai/studio/audio/voice-agents/tools.md): Give the agent web search, document libraries, connectors, and your own client-side functions — with fillers.
- [Webhooks](https://docs.mistral.ai/studio/audio/voice-agents/webhooks.md): Receive signed call lifecycle events — call started, ended, failed — on your endpoint.
- [Batch Processing](https://docs.mistral.ai/studio/batch-processing.md): Batching allows you to run asynchronous inference on large inputs in parallel, reducing compute costs while running large workloads at a 50% discount.
- [Human-in-the-loop](https://docs.mistral.ai/studio/connectors/confirmation.md): Some tool calls, like sending an email or modifying data, are safer with human approval.
- [Using Connectors in conversations](https://docs.mistral.ai/studio/connectors/conversations.md): After a Connector is registered and authenticated (if required), you can attach it to any conversation.
- [Debug Connectors](https://docs.mistral.ai/studio/connectors/debugger.md): The Connectors Debugger helps you test an MCP Connector server before you use it with Studio.
- [Managing Connectors](https://docs.mistral.ai/studio/connectors/management.md): Before you can use a Connector in conversations or call its tools, you need to register it.
- [Connectors](https://docs.mistral.ai/studio/connectors.md): Connectors are a Public Preview feature.
- [Google](https://docs.mistral.ai/studio/connectors/providers/google.md): How Vibe by Mistral AI uses Google APIs — scopes, data access, and privacy details for Gmail, Google Calendar, Google Drive, Google BigQuery, and Google Workspace.
- [Direct tool calling with Connectors](https://docs.mistral.ai/studio/connectors/tool_calling.md): The call_tool method lets you call a specific MCP tool on a Connector directly, without starting a conversation or involving the model.
- [Predicted outputs](https://docs.mistral.ai/studio/conversations/advanced/predicted-outputs.md): Predicted Outputs optimizes response time by leveraging known or predictable content.
- [Prompt caching](https://docs.mistral.ai/studio/conversations/advanced/prompt-caching.md): Prompt caching lets you reuse previously computed prompt tokens when requests share the same prefix, meaning the same beginning of the prompt.
- [Chat completions](https://docs.mistral.ai/studio/conversations/chat-completion.md): Large language models (LLMs) are AI systems that generate text and engage in conversational interactions.
- [Using prompts in chat completions](https://docs.mistral.ai/studio/conversations/chat-completion/prompt-registry.md): The Prompt Registry lets you store, version, and manage prompts centrally in Studio.
- [Prompting](https://docs.mistral.ai/studio/conversations/chat-completion/prompting.md): When you first start using Mistral models, your initial interaction will revolve around prompts.
- [Citations & References](https://docs.mistral.ai/studio/conversations/citations.md): Citations enable models to ground their responses and provide references, making them a powerful feature for Retrieval-Augmented Generation (RAG) and agentic applications.
- [Function Calling](https://docs.mistral.ai/studio/conversations/function-calling.md): Function calling, under the Tool Calling umbrella, allows Mistral models to connect to external local tools.
- [Moderation & Guardrailing](https://docs.mistral.ai/studio/conversations/moderation.md): When deploying LLMs in production, different verticals may require different levels of guardrailing.
- [Reasoning](https://docs.mistral.ai/studio/conversations/reasoning.md): Reasoning is the next step of CoT (Chain of Thought), naturally used to describe the logical steps generated by the model before reaching a conclusion.
- [Custom](https://docs.mistral.ai/studio/conversations/structured-output/custom.md): Custom Structured Outputs allow you to ensure the model provides an answer in a very specific JSON format by supplying a clear JSON schema.
- [JSON Mode](https://docs.mistral.ai/studio/conversations/structured-output/json_mode.md): Users have the option to set response_format to {type: json_object} to enable JSON mode.
- [Structured Outputs](https://docs.mistral.ai/studio/conversations/structured-output.md): When utilizing LLMs as agents or steps within a lengthy process, chain, or pipeline, it is often necessary for the outputs to adhere to a specific structured format.
- [Vision](https://docs.mistral.ai/studio/conversations/vision.md): Vision capabilities enable models to analyze images and provide insights based on visual content in addition to text.
- [Document Annotations](https://docs.mistral.ai/studio/document-processing/annotations.md): In addition to the basic OCR functionality, Mistral Document AI API adds the annotations functionality, which allows you to extract information in a structured json-format that you provide.
- [OCR Processor](https://docs.mistral.ai/studio/document-processing/basic_ocr.md): Use the Document AI OCR processor to extract text and structured content from PDF documents and images.
- [Document QnA](https://docs.mistral.ai/studio/document-processing/document_qna.md): The Document QnA capability combines OCR with large language model capabilities to enable natural language interaction with document content.
- [Document Processing](https://docs.mistral.ai/studio/document-processing/overview.md): Mistral Document AI offers enterprise-level document processing, combining OCR technology with advanced structured data extraction.
- [Code Embeddings](https://docs.mistral.ai/studio/knowledge-rag/embeddings/code_embeddings.md): Embeddings are at the core of multiple enterprise use cases, such as retrieval systems, clustering, code analytics, classification, and a variety of search applications.
- [Embeddings](https://docs.mistral.ai/studio/knowledge-rag/embeddings.md): Embeddings are vector representations of text that capture the semantic meaning of paragraphs through their position in a high-dimensional vector space.
- [Text Embeddings](https://docs.mistral.ai/studio/knowledge-rag/embeddings/text_embeddings.md): Embeddings are at the core of multiple enterprise use cases, such as retrieval systems, clustering, code analytics, classification, and a variety of search applications.
- [Use context objects](https://docs.mistral.ai/studio/observability/evaluations/advanced-guides/context-objects.md): Tasks, scorers, and metadata callbacks receive their inputs through context objects: typed Pydantic models that bundle all available data into a single parameter.
- [Iterate locally](https://docs.mistral.ai/studio/observability/evaluations/advanced-guides/local-mode.md): Set local=True to run evaluations without uploading results to Studio.
- [Multiple evaluators](https://docs.mistral.ai/studio/observability/evaluations/advanced-guides/multiple-evaluators.md): A run can hold any number of evaluators — pass them as a list to evaluators.
- [Reduce variance with multiple generations](https://docs.mistral.ai/studio/observability/evaluations/advanced-guides/num-generations.md): When your task is non-deterministic (for example, temperature > 0), a single generation per input may not give you a reliable picture.
- [Rescore persisted runs](https://docs.mistral.ai/studio/observability/evaluations/advanced-guides/rescoring.md): An expensive task execution should be produced once and scored as many times as you like. evaluation.rescore() scores the outputs already persisted in Studio without re-running the task — useful when the generations a...
- [Retry failed records](https://docs.mistral.ai/studio/observability/evaluations/advanced-guides/retry-failed-records.md): Evaluations can fail partially: a transient API error, a rate limit, or a bug in your scorer.
- [Use run-level evaluators](https://docs.mistral.ai/studio/observability/evaluations/advanced-guides/run-evaluators.md): While regular evaluators score each record individually, run evaluators operate on the full set of results after all records have been processed.
- [API reference](https://docs.mistral.ai/studio/observability/evaluations/api-reference.md): Reference for the Evaluation SDK's core types and the main evaluation.run() entry point.
- [Datasets](https://docs.mistral.ai/studio/observability/evaluations/datasets.md): A dataset is the set of test cases that drives an offline evaluation.
- [Evaluators](https://docs.mistral.ai/studio/observability/evaluations/evaluators.md): An evaluator scores each record of a run.
- [Set goals](https://docs.mistral.ai/studio/observability/evaluations/goals.md): Goals let you define pass/fail criteria on evaluator scores.
- [Optimize prompts and parameters](https://docs.mistral.ai/studio/observability/evaluations/optimization.md): Optimization turns an evaluation into a search: instead of measuring a single prompt or set of system params, the SDK automatically explores variations of it and returns the best one it finds — scored against the same...
- [Offline evaluations](https://docs.mistral.ai/studio/observability/evaluations.md): The Evaluation SDK (mistralai-evaluations) lets you run offline evaluations on your LLM pipelines in a few lines of Python.
- [Configure statistics](https://docs.mistral.ai/studio/observability/evaluations/statistics.md): Statistics are the run-level values an evaluator exposes from its per-record scores: an average, a total, a percentile, or a classification metric such as F1.
- [Configure system params](https://docs.mistral.ai/studio/observability/evaluations/system-params.md): When you run an LLM pipeline, many parameters influence the output: the model, temperature, system prompt, tool definitions, retrieval settings, and more.
- [Workflow evaluation plugin API reference](https://docs.mistral.ai/studio/observability/evaluations/workflows-plugin/api-reference.md): Reference for evaluation.run(), its result, and the plugin's imports.
- [Building blocks](https://docs.mistral.ai/studio/observability/evaluations/workflows-plugin/building-blocks.md): When your workflow already manages its own fan-out, for example with custom child workflow orchestration, call the steps of evaluation.run() yourself instead.
- [Decorators](https://docs.mistral.ai/studio/observability/evaluations/workflows-plugin/decorators.md): Every function you pass to the plugin runs as a Temporal activity.
- [Optimize prompts and parameters in a workflow](https://docs.mistral.ai/studio/observability/evaluations/workflows-plugin/optimization.md): evaluation.optimize() runs the optimizer inside a workflow.
- [Workflow evaluation plugin](https://docs.mistral.ai/studio/observability/evaluations/workflows-plugin.md): The Workflow evaluation plugin (mistralai-workflows-plugins-evaluations) runs offline evaluations inside Mistral Workflows.
- [Rescore persisted runs in a workflow](https://docs.mistral.ai/studio/observability/evaluations/workflows-plugin/rescoring.md): evaluation.rescore() scores the outputs already persisted in Studio without re-running the task, with your scorers running as activities.
- [Task modes](https://docs.mistral.ai/studio/observability/evaluations/workflows-plugin/task-modes.md): The task parameter of evaluation.run() accepts three kinds of values.
- [Observability](https://docs.mistral.ai/studio/observability.md): Observability is in Private Preview and is available for Enterprise-tier organizations only.
- [Data redaction](https://docs.mistral.ai/studio/observability/traces/data-redaction.md): Spans capture the input and output of each operation, including prompts, responses, and tool call arguments and results.
- [Explore traces](https://docs.mistral.ai/studio/observability/traces/explorer.md): The Trace Explorer is where you search, filter, and inspect every trace flowing through your AI applications.
- [Distributed tracing](https://docs.mistral.ai/studio/observability/traces.md): Every request in your AI application is captured as a trace: a tree of spans representing every step in the execution chain.
- [Send traces](https://docs.mistral.ai/studio/observability/traces/send-traces.md): This page covers how to instrument each supported data source so it sends OpenTelemetry traces to Mistral.
- [Studio](https://docs.mistral.ai/studio.md): Studio is Mistral's developer platform for AI applications.
- [Moderation & Guardrailing](https://docs.mistral.ai/studio/safety-moderation.md): When deploying LLMs in production, different verticals may require different levels of guardrailing.
- [Agentic Search](https://docs.mistral.ai/studio/search/agentic-search.md): Agentic Search helps agents answer questions when the answer is not in the first chunk returned by retrieval.
- [Libraries](https://docs.mistral.ai/studio/search/libraries.md): This page covers how to create Libraries, upload documents, index web pages, check processing status, and control access.
- [Document model](https://docs.mistral.ai/studio/search/search-toolkit/concepts/document-model.md): How Search Toolkit represents documents and chunks, and the deterministic identity that ties them together.
- [Embedding model](https://docs.mistral.ai/studio/search/search-toolkit/concepts/embedding-model.md): How Search Toolkit describes embedding vector dimensions, data type, and distance metric separately from the embedder that produces them.
- [Embedders](https://docs.mistral.ai/studio/search/search-toolkit/ingestion/embedders.md): Convert text into vector embeddings for semantic search.
- [Chunk enrichers](https://docs.mistral.ai/studio/search/search-toolkit/ingestion/enrichers.md): Add custom metadata to chunks during ingestion.
- [Document extractors](https://docs.mistral.ai/studio/search/search-toolkit/ingestion/extractors.md): Extract content from File objects into Document objects for processing.
- [File loaders](https://docs.mistral.ai/studio/search/search-toolkit/ingestion/loaders.md): Load files from various sources into File objects for ingestion.
- [Ingestion](https://docs.mistral.ai/studio/search/search-toolkit/ingestion.md): Ingestion transforms raw documents into searchable chunks indexed in a vector store.
- [Text splitters](https://docs.mistral.ai/studio/search/search-toolkit/ingestion/splitters.md): Divide documents into retrievable DocumentChunk objects for indexing.
- [Search Toolkit](https://docs.mistral.ai/studio/search/search-toolkit.md): Search Toolkit is a Python framework for building information retrieval (IR) systems.
- [Quickstart](https://docs.mistral.ai/studio/search/search-toolkit/quickstart.md): Build a RAG pipeline in 5 minutes: ingest documents into a vector store, then search them.
- [Retrieval](https://docs.mistral.ai/studio/search/search-toolkit/retrieval.md): Retrieval finds relevant chunks for a given user query.
- [Query preprocessing](https://docs.mistral.ai/studio/search/search-toolkit/retrieval/preprocessing.md): Improve user queries before retrieval for better results.
- [Rerankers](https://docs.mistral.ai/studio/search/search-toolkit/retrieval/rerankers.md): Re-score and refine retrieved results for better ranking.
- [Retrievers](https://docs.mistral.ai/studio/search/search-toolkit/retrieval/retrievers.md): Search your index using vector, keyword, or hybrid strategies.
- [Semantic cache](https://docs.mistral.ai/studio/search/search-toolkit/retrieval/semantic-cache.md): Cache retrieval results by query similarity to reduce latency.
- [Custom vector stores](https://docs.mistral.ai/studio/search/search-toolkit/search-index/custom-vector-stores.md): Implement your own storage backend for Search Toolkit.
- [Search index](https://docs.mistral.ai/studio/search/search-toolkit/search-index.md): Persist processed chunks and enable efficient search across your document collection.
- [Postgres](https://docs.mistral.ai/studio/search/search-toolkit/search-index/postgres.md): Declare collections, manage the database, and run dense and hybrid search on PostgreSQL with pgvector.
- [Qdrant](https://docs.mistral.ai/studio/search/search-toolkit/search-index/qdrant.md): Community-managed Qdrant search index backend plugin.
- [Anatomy of a Vespa application](https://docs.mistral.ai/studio/search/search-toolkit/search-index/vespa/anatomy.md): This page explains the concepts that make up a Vespa application: the application package, schemas, fields, ranking profiles, and query profiles.
- [CLI reference](https://docs.mistral.ai/studio/search/search-toolkit/search-index/vespa/cli.md): The mistral-vespa CLI manages the Vespa application lifecycle, from generating migrations to deployment and testing.
- [Local development](https://docs.mistral.ai/studio/search/search-toolkit/search-index/vespa/local-development.md): This guide covers the full local development loop: create a schema, deploy locally, feed and query documents, then iterate with new migrations.
- [Migration helpers reference](https://docs.mistral.ai/studio/search/search-toolkit/search-index/vespa/migration-helpers.md): Reference for the helpers used inside migrations.
- [Manage schema](https://docs.mistral.ai/studio/search/search-toolkit/search-index/vespa/migrations.md): Manage your Vespa application schemas through Python migrations.
- [Deploy and operate](https://docs.mistral.ai/studio/search/search-toolkit/search-index/vespa/operations.md): Production deployment and operation of Vespa applications.
- [Vespa](https://docs.mistral.ai/studio/search/search-toolkit/search-index/vespa.md): Define, deploy, and operate Vespa applications - schema management, migrations, ranking configuration, and production deployment.
- [Manage ranking](https://docs.mistral.ai/studio/search/search-toolkit/search-index/vespa/query-profiles.md): Query profiles let you control ranking at query time without modifying your schema.
- [Secrets Manager](https://docs.mistral.ai/studio/secrets.md): Store credentials in Secrets Manager and reference them from your managed deployments without hardcoding.
- [Basics](https://docs.mistral.ai/studio/workflows/building-workflows/activities/basics.md): Activities are the units of work in a workflow.
- [Local activities](https://docs.mistral.ai/studio/workflows/building-workflows/activities/local_activities.md): Local activities run directly in the workflow worker process, skipping the regular platform scheduling hop.
- [Activities](https://docs.mistral.ai/studio/workflows/building-workflows/activities.md): This page covers the practical patterns for building activities.
- [Sticky worker sessions](https://docs.mistral.ai/studio/workflows/building-workflows/activities/sticky_worker_sessions.md): Sticky worker sessions route multiple activities to the same worker instance, letting you reuse resources and maintain state across activity calls.
- [Compression](https://docs.mistral.ai/studio/workflows/building-workflows/compression.md): Compress all payloads (workflow inputs, activity I/O, signal data) before they leave your worker.
- [Connectors in workflows](https://docs.mistral.ai/studio/workflows/building-workflows/connectors.md): Use Connectors inside a workflow to call external services, such as GitHub, Notion, Slack, and Outlook, without managing credentials yourself.
- [Consuming Streaming Events](https://docs.mistral.ai/studio/workflows/building-workflows/consuming_events.md): As a workflow runs, it emits events — execution started, activity completed, tokens generated.
- [Continue-As-New](https://docs.mistral.ai/studio/workflows/building-workflows/continue_as_new.md): Continue-As-New resets the workflow's event history while carrying forward its state, enabling indefinitely long-running workflows without hitting history size limits.
- [Dependency Injection](https://docs.mistral.ai/studio/workflows/building-workflows/dependency_injection.md): Dependency injection provides a clean way to pass shared resources (database connections, API clients, configuration) into activities without constructing them manually inside each call.
- [Durable agents](https://docs.mistral.ai/studio/workflows/building-workflows/durable_agents.md): A durable agent is an LLM agent whose loop (model calls, tool use, and handoffs) runs inside a workflow, so its state survives crashes and restarts.
- [Encryption](https://docs.mistral.ai/studio/workflows/building-workflows/encryption.md): Encrypt all payloads (workflow inputs, activity I/O, signal data) before they leave your worker.
- [Mistral client](https://docs.mistral.ai/studio/workflows/building-workflows/mistral_client.md): get_mistral_client() returns a standard Mistral SDK client pre-configured with workflow-aware hooks.
- [On-behalf-of workflows](https://docs.mistral.ai/studio/workflows/building-workflows/on_behalf_of.md): On-behalf-of workflows require a hardened deployment.
- [Payload offloading](https://docs.mistral.ai/studio/workflows/building-workflows/payload_offloading.md): To keep the orchestration layer fast and predictable for every workflow on the platform, Mistral Workflows enforces a 2MB limit on workflow inputs, activity inputs, and activity outputs.
- [Plugins](https://docs.mistral.ai/studio/workflows/building-workflows/plugins.md): Mistral Workflows plugins are standard Python packages that expose reusable workflows, activities, and dependencies under the mistralai.workflows.plugins namespace.
- [Workflow scheduling](https://docs.mistral.ai/studio/workflows/building-workflows/scheduling.md): Run workflows on a recurring cron, interval, or calendar schedule.
- [Search keys](https://docs.mistral.ai/studio/workflows/building-workflows/search_keys.md): Search keys are key/value pairs attached to an execution so it's easier to find later.
- [Streaming Events](https://docs.mistral.ai/studio/workflows/building-workflows/streaming.md): Stream events in real-time from your workflows and activities to power live UIs, progress indicators, and token-by-token LLM responses.
- [Child workflows](https://docs.mistral.ai/studio/workflows/building-workflows/sub_workflows.md): A workflow can execute other workflows as child workflows, enabling hierarchical orchestration patterns.
- [Waiting for Conditions](https://docs.mistral.ai/studio/workflows/building-workflows/waiting_for_conditions.md): Workflows can pause execution until a specific condition is met, enabling event-driven patterns like human-in-the-loop and approval flows.
- [Workflows Exception](https://docs.mistral.ai/studio/workflows/building-workflows/workflow_exception.md): Workflows provides a structured exception class, WorkflowsException, for consistent error handling across workflows and activities.
- [Workflow Tags](https://docs.mistral.ai/studio/workflows/building-workflows/workflow_tags.md): Workflows support searchable metadata tags for filtering and discovery.
- [Determinism](https://docs.mistral.ai/studio/workflows/building-workflows/workflows/determinism.md): Workflows must be deterministic: given the same inputs, they must always produce the same
- [Workflows](https://docs.mistral.ai/studio/workflows/building-workflows/workflows.md): This page covers the practical patterns for building workflows.
- [Cookbook examples](https://docs.mistral.ai/studio/workflows/getting-started/cookbook_examples.md): Use the Workflows cookbook templates when you want to start with complete examples.
- [Activities](https://docs.mistral.ai/studio/workflows/getting-started/core_concepts/activities.md): An activity is where the actual work happens: sending an email, calling an external API, running a database query, processing a file, generating a response from an AI model.
- [Deployments](https://docs.mistral.ai/studio/workflows/getting-started/core_concepts/deployments.md): A deployment is a named group of workers that owns a set of workflow definitions and receives all executions for those definitions.
- [Events](https://docs.mistral.ai/studio/workflows/getting-started/core_concepts/events.md): Every significant action in a workflow's lifetime produces an event: the workflow started, an activity was scheduled, an activity completed, a signal was received, the workflow finished.
- [Executions](https://docs.mistral.ai/studio/workflows/getting-started/core_concepts/executions.md): An execution is a single invocation of a workflow.
- [Core concepts](https://docs.mistral.ai/studio/workflows/getting-started/core_concepts.md): This section explains the building blocks of Mistral Workflows and how they fit together.
- [Workers](https://docs.mistral.ai/studio/workflows/getting-started/core_concepts/workers.md): A worker is the process that actually runs your code.
- [Workflows](https://docs.mistral.ai/studio/workflows/getting-started/core_concepts/workflows.md): A workflow is the composition layer of your application: it defines what to do and when, while activities define how.
- [Installation](https://docs.mistral.ai/studio/workflows/getting-started/installation.md): This guide will walk you through setting up Workflows and verifying your installation.
- [Overview](https://docs.mistral.ai/studio/workflows/getting-started/overview.md): Workflows is in Public Preview.
- [Your first workflow](https://docs.mistral.ai/studio/workflows/getting-started/your_first_workflow.md): This guide walks you through creating a workflow that executes a single activity.
- [Canvas](https://docs.mistral.ai/studio/workflows/interacting-with-workflows/conversational_workflows/canvas.md): Canvas is the rich-content surface in the chat interface: markdown documents, code, diagrams, slides, and interactive components.
- [Forms and confirmations](https://docs.mistral.ai/studio/workflows/interacting-with-workflows/conversational_workflows/forms_and_confirmations.md): Conversational workflows can ask the user for structured input — typed fields with validation, single- or multi-choice options, file uploads, accept/decline confirmations.
- [Conversational Workflows](https://docs.mistral.ai/studio/workflows/interacting-with-workflows/conversational_workflows.md): Conversational workflows are meant to be integrated in conversation interfaces, allowing a user to trigger a workflow, interact with it by providing inputs during its execution, and follow its progress.
- [Progress tracking](https://docs.mistral.ai/studio/workflows/interacting-with-workflows/conversational_workflows/progress_tracking.md): Display a checklist of steps with real-time status updates using TodoList.
- [Publish in Vibe](https://docs.mistral.ai/studio/workflows/interacting-with-workflows/conversational_workflows/publish_in_vibe.md): To publish a conversational workflow as an assistant that surfaces in Vibe (specifically in Vibe Work, the web and mobile chat UI), your workflow must return a ChatAssistantWorkflowOutput.
- [Tool UI](https://docs.mistral.ai/studio/workflows/interacting-with-workflows/conversational_workflows/tool_ui.md): Render rich, interactive UI components in the Vibe Work chat interface and visualize tool execution with structured status feedback.
- [Queries](https://docs.mistral.ai/studio/workflows/interacting-with-workflows/queries.md): Queries allow external systems to read the current state of a running workflow synchronously.
- [Signals](https://docs.mistral.ai/studio/workflows/interacting-with-workflows/signals.md): Signals allow external systems to send messages to running workflows asynchronously.
- [Updates](https://docs.mistral.ai/studio/workflows/interacting-with-workflows/updates.md): Updates allow external systems to modify workflow state and receive a response.
- [Concurrency Patterns](https://docs.mistral.ai/studio/workflows/managing-workflows-in-production/concurrency.md): _Process thousands of items efficiently with Mistral Workflows' parallel execution patterns_
- [Deployments](https://docs.mistral.ai/studio/workflows/managing-workflows-in-production/deployments.md): A deployment is a named group of workers that owns a set of workflow definitions and receives all executions for those definitions.
- [API Error Codes](https://docs.mistral.ai/studio/workflows/managing-workflows-in-production/error_codes.md): When a request to the Workflows API fails, the response includes a structured error code in WF_XXXX format:
- [Execution Context](https://docs.mistral.ai/studio/workflows/managing-workflows-in-production/execution_context.md): Access runtime information about the current workflow execution from within workflow code.
- [Hardened deployments](https://docs.mistral.ai/studio/workflows/managing-workflows-in-production/hardened_deployments.md): A hardened deployment is a deployment that restricts workflow registration to authorized users and service accounts.
- [Getting started with managed deployments](https://docs.mistral.ai/studio/workflows/managing-workflows-in-production/managed-deployments/getting-started.md): Scaffold the Workflows starter app, push it to GitHub, and let Mistral Cloud build and run the worker for you.
- [How it works](https://docs.mistral.ai/studio/workflows/managing-workflows-in-production/managed-deployments/how-it-works.md): What Mistral Cloud does to your worker container: service account, Dockerfile entrypoint, and injected environment variables.
- [Managed deployments](https://docs.mistral.ai/studio/workflows/managing-workflows-in-production/managed-deployments.md): Deploy a workflow worker from a GitHub repository.
- [Quotas](https://docs.mistral.ai/studio/workflows/managing-workflows-in-production/managed-deployments/quotas.md): Concurrent managed deployment caps per plan, and instance sizing.
- [Secrets](https://docs.mistral.ai/studio/workflows/managing-workflows-in-production/managed-deployments/secrets.md): Bind workspace secrets to a managed deployment, available at build and at runtime.
- [Troubleshooting](https://docs.mistral.ai/studio/workflows/managing-workflows-in-production/managed-deployments/troubleshooting.md): Common managed deployment failures, and how to reproduce a build locally.
- [Rate Limiting](https://docs.mistral.ai/studio/workflows/managing-workflows-in-production/rate_limiting.md): Rate limiting is a crucial aspect of workflow management that helps control resource consumption and prevent any single workflow or activity from monopolizing shared resources.
- [Resetting Workflows](https://docs.mistral.ai/studio/workflows/managing-workflows-in-production/reset_workflow.md): Reset a workflow execution to restart it from a specific point in its event history.
- [Observability](https://docs.mistral.ai/studio/workflows/observability.md): This guide covers how worker telemetry (logs, traces, and metrics) is exported, and how to use OpenTelemetry traces for execution-level diagnostics.
- [Workflow visualizer](https://docs.mistral.ai/studio/workflows/visualizer.md): The workflow visualizer displays your workflow as a graph in Studio.

## Vibe

- [Agents](https://docs.mistral.ai/vibe/chat-legacy/agents.md): Agents are a legacy Chat feature.
- [Code Interpreter](https://docs.mistral.ai/vibe/chat-legacy/code-interpreter.md): Code Interpreter is a legacy Chat feature.
- [Deep Research](https://docs.mistral.ai/vibe/chat-legacy/deep-research.md): Deep Research is a legacy Chat feature.
- [Memories](https://docs.mistral.ai/vibe/chat-legacy/memories.md): Memories is a legacy Chat feature.
- [Think mode](https://docs.mistral.ai/vibe/chat-legacy/think-mode.md): Think mode is a legacy Chat feature.
- [Choose Work or Code](https://docs.mistral.ai/vibe/choose-chat-work-code.md): A closer look at each Vibe mode: when to use it, what it does, and where to start.
- [Choose CLI, VS Code, or web sessions](https://docs.mistral.ai/vibe/code/choose-cli-vscode-web-sessions.md): One agent, three surfaces. Vibe Code runs as the CLI, the VS Code extension, and Vibe Code Web.
- [Admin config](https://docs.mistral.ai/vibe/code/cli/admin-config.md): Admin config, also called managed config, lets Organization and Workspace administrators distribute settings for the Vibe Code CLI and the VS Code extension to all users.
- [Agents](https://docs.mistral.ai/vibe/code/cli/agents.md): Agents are configuration overrides applied on top of the global config.
- [API keys and profiles](https://docs.mistral.ai/vibe/code/cli/api-keys-profiles.md): Vibe Code CLI needs a Mistral API key to call hosted models.
- [Commands and shortcuts](https://docs.mistral.ai/vibe/code/cli/commands-shortcuts.md): Use slash commands and keyboard shortcuts to control a Vibe session without leaving the terminal.
- [Configuration reference](https://docs.mistral.ai/vibe/code/cli/configuration-reference.md): An exhaustive index of every key accepted by config.toml and the agent-definition files.
- [Configuration](https://docs.mistral.ai/vibe/code/cli/configuration.md): The Vibe Code CLI is configured through a config.toml file.
- [Connectors](https://docs.mistral.ai/vibe/code/cli/connectors.md): Connectors give the Vibe Code CLI access to supported first-party and third-party services from inside a coding session.
- [Hooks](https://docs.mistral.ai/vibe/code/cli/hooks.md): Hooks wire arbitrary shell commands into the Vibe Code CLI's lifecycle to gate, audit, or rewrite agent behavior.
- [Install and setup](https://docs.mistral.ai/vibe/code/cli/install-setup.md): Install the Vibe CLI, run vibe in your project, and complete the setup prompt.
- [MCP servers](https://docs.mistral.ai/vibe/code/cli/mcp-servers.md): MCP servers extend the Vibe Code CLI with external tools through the Model Context Protocol.
- [Using offline models](https://docs.mistral.ai/vibe/code/cli/offline-models.md): The Vibe Code CLI supports any model served behind an OpenAI-compatible API.
- [Plugins](https://docs.mistral.ai/vibe/code/cli/plugins.md): Plugin support is experimental and under active development.
- [Skills](https://docs.mistral.ai/vibe/code/cli/skills.md): Skills are reusable instruction sets that extend Vibe Code with new workflows and custom slash commands.
- [Teleport from CLI to web](https://docs.mistral.ai/vibe/code/cli/teleport-cli-web.md): Teleport moves a CLI session into a Vibe Code Web sandbox so the task can keep running without your terminal staying open.
- [Terminal support](https://docs.mistral.ai/vibe/code/cli/terminal-support.md): The Vibe CLI runs in any modern terminal emulator, but terminals don't all support the same features.
- [Work with the CLI](https://docs.mistral.ai/vibe/code/cli/work-with-cli.md): Once the CLI is installed and your API key is set, you're ready for day-to-day work.
- [Vibe Code](https://docs.mistral.ai/vibe/code/overview.md): Vibe Code is Vibe's coding mode.
- [Safety, approvals, and permissions](https://docs.mistral.ai/vibe/code/safety-approvals-permissions.md): Vibe Code can read files, edit code, run shell commands, and call external tools on your behalf.
- [Use Vibe in other IDEs](https://docs.mistral.ai/vibe/code/use-vibe-in-other-ides.md): Vibe implements the Agent Client Protocol (ACP) and is published in the ACP registry.
- [Get started](https://docs.mistral.ai/vibe/code/vibe-code-web/get-started.md): Vibe Code Web is the control plane for remote coding agents.
- [GitHub repositories](https://docs.mistral.ai/vibe/code/vibe-code-web/github-repositories-permissions.md): Vibe Code Web works on GitHub repositories from a managed cloud sandbox.
- [Limits and lifecycle](https://docs.mistral.ai/vibe/code/vibe-code-web/limits-and-lifecycle.md): A Vibe Code Web session is one run of the Vibe Code agent against a GitHub repository in a managed cloud sandbox.
- [Projects](https://docs.mistral.ai/vibe/code/vibe-code-web/projects.md): A project is the workspace Vibe Code Web organizes everything around: one or more GitHub repositories, the sessions you run against them, and the history of past runs.
- [Sandbox environment](https://docs.mistral.ai/vibe/code/vibe-code-web/sandbox-environment.md): Each Vibe Code Web session runs in an isolated cloud sandbox provisioned for that run.
- [Security](https://docs.mistral.ai/vibe/code/vibe-code-web/security.md): This page describes the security model for autonomous agent runs in Vibe Code Web: how to handle untrusted content the agent reads, and how session data is retained.
- [Sessions](https://docs.mistral.ai/vibe/code/vibe-code-web/sessions.md): A session is one run of the Vibe Code agent against a project: a sandbox provisioned for the task, the agent loop that does the work, and the branch or pull request it leaves behind.
- [Vibe Code Web for Slack: Quickstart](https://docs.mistral.ai/vibe/code/vibe-code-web/slack-integration.md): AI-generated content: This app uses AI to generate responses and code.
- [Agents](https://docs.mistral.ai/vibe/code/vs-code-extension/agents.md): Agents are configuration overrides applied on top of the global config.
- [Commands and slash commands](https://docs.mistral.ai/vibe/code/vs-code-extension/commands-slash-commands.md): Use slash commands and keyboard shortcuts to control Vibe Code from inside the editor.
- [Install and authenticate](https://docs.mistral.ai/vibe/code/vs-code-extension/install-authenticate.md): Install Mistral Vibe for VS Code, open the Vibe panel, and sign in.
- [Migration from Mistral Code Enterprise](https://docs.mistral.ai/vibe/code/vs-code-extension/migration-mistral-code-enterprise.md): This page covers what to expect when moving from the Mistral Code extension to Mistral Vibe for VS Code.
- [Extension settings](https://docs.mistral.ai/vibe/code/vs-code-extension/settings.md): A dedicated settings panel for the VS Code extension is in progress.
- [Work with the extension](https://docs.mistral.ai/vibe/code/vs-code-extension/work-with-extension.md): Use the VS Code extension when you want Vibe Code in the same workspace as your files, selections, diffs, and source control, with no context switch and no extra terminal.
- [Vibe](https://docs.mistral.ai/vibe.md): Vibe is Mistral's unified agent for productivity and coding tasks, built for professional use and available across web, mobile, your code editor, and your terminal.
- [Code Interpreter](https://docs.mistral.ai/vibe/work/code-interpreter.md): Code Interpreter is available on paid plans only.
- [MCP Connectors](https://docs.mistral.ai/vibe/work/connectors/mcp-connectors.md): Beyond our featured Connectors, you can connect Work to third-party and custom services built on the Model Context Protocol (MCP).
- [Connect tools with Connectors](https://docs.mistral.ai/vibe/work/connectors.md): Connectors are secure bridges between Work and your external tools and data sources.
- [Set custom instructions](https://docs.mistral.ai/vibe/work/custom-instructions.md): Custom instructions let you define persistent preferences that shape how Work responds across every task.
- [Work with Files and Canvas](https://docs.mistral.ai/vibe/work/files-and-canvas.md): Files bring source material into a task. Canvas is the built-in editor where Work produces, previews, and refines outputs.
- [Get started with Work](https://docs.mistral.ai/vibe/work/get-started.md): Vibe Work is Vibe's productivity mode for delegating complex, multi-step tasks across your apps and tools.
- [Generate images](https://docs.mistral.ai/vibe/work/image-generation.md): Work can generate and edit images directly in the chat interface.
- [Knowledge Base](https://docs.mistral.ai/vibe/work/knowledge.md): Knowledge Base saves what matters from your conversations so Vibe can recall it later.
- [Add context with Libraries](https://docs.mistral.ai/vibe/work/libraries.md): Libraries are persistent knowledge bases you build from your own documents and web pages.
- [Mini apps](https://docs.mistral.ai/vibe/work/mini-apps.md): Vibe Work can build interactive mini apps: self-contained React applications that run directly in the Canvas panel.
- [Group tasks with Projects](https://docs.mistral.ai/vibe/work/projects.md): Projects let you group chats into distinct, manageable collections — and give each group its own tone, instructions, and shared files.
- [Safety and approvals](https://docs.mistral.ai/vibe/work/safety-and-approvals.md): Work is built for delegation with supervision.
- [Schedule tasks](https://docs.mistral.ai/vibe/work/scheduled-tasks.md): Public Preview. Scheduled tasks are available in Public Preview.
- [Reuse work with Skills](https://docs.mistral.ai/vibe/work/skills.md): Skills are reusable instructions and resources that give Work specialized capabilities for specific tasks.
- [Spreadsheets](https://docs.mistral.ai/vibe/work/spreadsheets.md): Spreadsheets are available on paid plans only, until further notice.
- [Switch Organization or Workspace](https://docs.mistral.ai/vibe/work/switch-organization-workspace.md): Your Mistral account belongs to one or more Organizations, and each Organization contains one or more Workspaces.
- [Use Voice mode](https://docs.mistral.ai/vibe/work/voice-mode.md): Voice mode lets you start a task in Work using your voice instead of typing.
- [Search the web](https://docs.mistral.ai/vibe/work/web-search-open-url.md): Work can browse the internet in real time and read specific web pages as part of a task. Web search answers questions with up-to-date, sourced information. Open URL lets Work read and analyze a page you already have t...
- [Run Workflows](https://docs.mistral.ai/vibe/work/workflows.md): If your team has built an internal automation (a data pipeline, a report generator, a ticketing flow), you can call it from Work by selecting it from the + menu.

## Optional

- [Full documentation as a single file](https://docs.mistral.ai/llms-full.txt): every page listed above concatenated, for offline context loading
- [OpenAPI specification](https://docs.mistral.ai/openapi.yaml): the public machine-readable Mistral API specification