Intégration MCP

Le serveur annonce actuellement 40 outils et maintient 70 outils canoniques appelables, chiffres lus sur le serveur en direct. La réponse tools/list du serveur connecté fait référence.

Minds fournit un serveur Model Context Protocol (MCP) qui permet aux assistants IA de créer des studies de recherche synthétiques, de mener des enquêtes multi-perspective et d'interagir directement avec des AI Minds. Cela permet une intégration avec ChatGPT, Claude, Codex, Gemini CLI, Cursor, VS Code et d'autres clients compatibles MCP.

Qu'est-ce que MCP ?

Le Model Context Protocol (MCP) est un standard ouvert qui permet aux assistants IA de se connecter de manière sécurisée à des outils et sources de données externes. Avec notre intégration MCP, vous pouvez :

Créer et gérer des AI Minds (experts, personas, digital twins)
Construire des studies de recherche avec des audiencees d'experts du domaine
Lancer des questions de type enquête sur plusieurs perspectives
Obtenir des analyses statistiques sur les réponses d'un study
Exporter des rapports PDF avec votre charte, des synthèses Markdown ou des données brutes
Organiser les Minds en audiencees réutilisables

URL du serveur MCP

https://getminds.ai/mcp

Pour les instructions par client, consultez le guide de configuration Minds MCP. Chaque outil canonique et la surface annoncée en direct sont décrits dans la référence des outils Minds MCP. Les auteurs d'agents doivent aussi lire le guide d'utilisation MCP pour les agents.

Authentification

Le serveur MCP prend en charge deux méthodes d'authentification :

OAuth 2.1 (recommandé) — Les clients comme ChatGPT, Claude, Codex et Gemini CLI gèrent cela automatiquement. Vous serez invité à vous connecter à votre compte Minds lors de la connexion. Le guide de configuration liste les autorisations demandées par un client.
API Key — Pour un accès programmatique, générez une clé dans Settings → API Keys (commence par minds_). Passez-la comme Authorization: Bearer minds_your_key.

Clients compatibles

Client	Widget	Auth	Configuration
ChatGPT	Selon le client	OAuth	Web : Plugins → Add → Create MCP App, ou Minds depuis l'annuaire des plugins ; application de bureau : Plugins → Add → Add MCP server (Streamable HTTP, puis Save → Restart → Authenticate)
Claude (claude.ai et Claude Desktop)	Selon le client	OAuth	Customize → Connectors → + → Add custom connector
Claude Code	Non	OAuth ou API key	`claude mcp add --transport http minds https://getminds.ai/mcp`, puis `/mcp` → Authenticate
Codex	Non	OAuth ou API key	`codex mcp add minds --url https://getminds.ai/mcp`, puis `codex mcp login minds`
Gemini CLI	Non	OAuth	Entrée `httpUrl` dans `~/.gemini/settings.json`
Cursor	Selon le client	OAuth	Entrée `url` dans `~/.cursor/mcp.json`
VS Code (GitHub Copilot)	Selon le client	OAuth	MCP: Add Server → HTTP, puis l'URL du serveur
Langdock	Non	OAuth	Integrations → Add MCP
Windsurf	Non	OAuth ou API key	Cascade … → Open MCP config file, puis une entrée `serverUrl`

Tous les clients voient les mêmes outils annoncés. Les instructions détaillées pour chaque client figurent dans le guide de configuration ; le guide ChatGPT présente chaque écran de ChatGPT.

Tous les clients reçoivent des liens cliquables pour ouvrir les résultats dans la webapp Minds, peu importe la prise en charge des widgets.

Outils annoncés et outils canoniques

tools/list renvoie 40 outils sélectionnés, conçus pour la recherche interactive courante. Le serveur maintient au total 70 outils canoniques appelables, y compris des outils de cycle de vie qu'il n'annonce pas. Les intégrations explicites peuvent appeler ces outils par leur nom canonique lorsque le client prend en charge les appels d'outils configurés ou directs. Cette séparation garde la surface par défaut du modèle ciblée tout en préservant l'intégralité du contrat de fonctionnalités pris en charge.

Des alias d'outils peuvent rester appelables pour la compatibilité, mais les agents doivent toujours utiliser les noms canoniques de la référence des outils.

Domaine	Outil	Description	Comportement
Audiences	`list_audiences`	List Audiences. Lists the authenticated user's Audiences, most recently updated first, one page at a time (limit, default 20, and offset; nextOffset continues), with Mind counts, sharing state, and links. includeMinds adds each Audience's member Minds. searchQuery filters by name (case-insensitive, all matches, paged). Each Audience's workspaceUrl is the authenticated workspace link, valid verbatim; sharedAudienceUrl, when present, is the public share link for recipients. addOnRequired marks a paid Marketplace Audience the user has not subscribed to: prefer the user's own Audiences, and only use it after the user subscribes.	Lecture seule
	`import_audience_sources`	Import Audience Research Sources. Imports supplied UTF-8 text, Markdown, CSV and JSON research files into account-owned storage. Accepts file contents rather than local paths. Identical file imports are safe to retry. Optional caller-reviewed grounding JSON binds distributions to existingFiles followed by files in sourceIdx order, returning a normalized snapshot and checksum for audience preview and creation. This operation creates no Audience or Minds and performs no web search or independent verification of supplied percentages.	Écriture
	`create_audience_from_brief`	Create a Grounded Audience from a Brief. Researches a free-text population brief and builds a new Audience of Minds with explicit profiles and exact segment allocation. A new brief returns an evidence-backed draft for review; draft.rework or the displayed review revises that draft before confirmation, and creation requires the user-accepted draft or quota snapshot and its checksum. Every new brief starts a separate draft and, once confirmed, a separate Audience. An Audience that already exists is changed in place by recalibrating it by its id, never rebuilt from a new brief. Supports reviewed respondent datasets. Creation is asynchronous and returns an operation ID. Audiences are private unless link sharing is enabled.	Écriture
	`get_audience_creation_progress`	Audience creation progress. Reads progress and never starts or retries anything: an Audience creation by operationId, with its members’ training; or an existing Audience by audienceId, with its member training and background research, or with recalibrationId one recalibration’s status, retained result or terminal error.	Lecture seule
	`render_audience_review`	Show Audience Creation. Shows source research, distributions, proposed Minds, the draft and training for one or more Audience operations, display only: the user confirms or asks for changes in chat. Multiple operations appear in independent tabs within one widget.	Lecture seule
	`get_audience_limits`	Get Audience Limits. Returns the Audience size ceilings for the authenticated account before an Audience is created: the per-Audience plan cap (including any team allowance), the custom-size maximum and the per-mode ceilings. Relevant whenever a size is named, "as many as possible" is asked for, or a creation mode is chosen.	Lecture seule
	`ask_audience`	Ask One Standalone Audience Question. Asks exactly one standalone question of one existing Audience. It creates a private Study for that Audience, starts asynchronous answers from its Minds, and returns the Study identifier and its links.	Écriture
	`export_audience`	Export Audience Brief. Exports an Audience brief, or with kind "validation_report" the report of its validations (overall score calculation, every KPI, per-question answer shares, provenance), through the same renderer used by the web app. Brief: Markdown, PDF, DOCX, PPTX. Validation report: Markdown, PDF, DOCX, XLSX. Binary artifacts are returned as base64.	Écriture
	`duplicate_audience`	Duplicate Audience. Copy an Audience with independent copies of its Minds and all they know.	Écriture
	`recalibrate_audience`	Change an Existing Audience. Changes an existing Audience in place: it keeps its id, its Minds keep theirs (except copies below), and no Mind is deleted. Covers composition and trait edits, a population pivot, adding members, an exact chart selection (`distributions`), new sources and going deeper (`deepen`), and re-runs its research grounding. Owner, or verified Minds staff with existing direct Audience admin collaboration or an active owner-issued support grant. Staff support is limited to composition amendments: no growth, new sources, Go deeper or shared-Mind waiver. Asynchronous: the result is an accepted operation with its id, not a completed change. Queues behind another edit of the Audience and waits for a Study run on its members. Members shared with other Audiences are replaced here by copies with new ids. Reusing requestKey with identical input retries safely.	Destructif
	`manage_audience_support`	Manage Audience Support Access. Manage owner-issued, expiring staff access to one Audience’s composition and receipts. Staff cannot self-grant. Revocation stops new work at checkpoints; completed changes remain. No membership, growth, deletion or billing changes.	Destructif
	`attach_study_audiences`	Attach Audiences to Study. Adds existing accessible Audiences without removing others, planning or starting research. Returns attached and alreadyAttached IDs; retries are safe. With link sharing on, the Audiences and their Minds become publicly readable. Access and plan eligibility are validated first.	Écriture
Studies	`list_studies`	List Studies. Lists the authenticated user's Studies, most recently updated first, one page at a time: limit (default 20) and offset, with nextOffset to continue. Each row carries the Study's Audiences with their Mind counts, stored message count, sharing state and links. searchQuery filters by name (case-insensitive, all matches, paged). Minds and per-question results come from the Study status.	Lecture seule
	`create_study`	Create a Study. Creates a Study workspace from existing Audiences or inline Audience configurations. It does not ask questions or run research, and follow-up research inside an existing Study needs no new Study.	Écriture
	`ask_study`	Ask One Standalone Question in a Study. Submits exactly one respondent-visible question in an existing Study: one standalone question, or one adaptive follow-up whose wording could not be known before earlier results.	Écriture
	`get_study_status`	Get Study Status. Returns Study progress, per-Audience results, original Mind responses, links and export status. In-flight questions report zero Minds answered until completion; partial per-Mind progress is not persisted. questionId selects one question; runId selects a confirmed run with question-level counters. Without either, results cover the entire Study. Values and labels are written in the Study display locale, returned as locale: the language of its confirmed plan, else the stored Study locale, else English.	Lecture seule
	`render_study_results`	Show Final Study Results. Shows the results of finished questions (summary, key findings, Audience breakdowns, Mind responses): one question, or a run's finished questions while the rest still run. Use it whenever the user asks about results. Requires exactly one of questionId or runId.	Lecture seule
	`export_study`	Export Study Results. Starts an asynchronous export of Study results and returns an export job ID. Supports executive briefs and full reports in PDF, DOCX, PPTX, or Markdown, plus raw data in CSV, XLS, or SPSS SAV.	Écriture
	`duplicate_study`	Duplicate Study. Copy a Study with all its questions and results over the same Audiences.	Écriture
	`export_heatmap`	Export Website Heatmap. Exports a completed website heatmap from a Study result, identified by the message ID reported with the completed result. Returns the same ZIP archive as the web app, including its unified-renderer PDF report, Markdown, images, and metadata.	Écriture
	`get_study_heatmap`	Read Study Asset Heatmap. Reads saved heatmaps of an assigned question asset. Optional assetKey selects its saved upload path or normalized URL. Returns assetHeatmaps without starting analysis, spending responses or visiting websites.	Lecture seule
	`start_study_heatmap`	Start Study Asset Heatmap. Starts attention analysis of an assigned image, video or website. Uses one response per Mind and needs a paid plan. assetKey reuses completed analysis of that asset; omitting it may replace earlier derived results. Websites are visited publicly. Explicit approval is required; question answers are preserved.	Destructif
	`create_study_plan`	Create Study Plan. Creates a non-executing draft in an existing Study with its Audiences already attached. The setup operation for two or more known questions: every question known now belongs in this one draft, in cohesive named modules. A one-question draft is only for genuinely standalone research or an adaptive follow-up. request designs questions; questions preserves a fixed instrument verbatim. Each Mind sees only its own earlier answers; askIf applies respondent-level skip logic. With multiple attachments, stimulus.questionAttachments assigns each question; [] means none. Saving bounded history may prune older revisions. Research has not started; the returned draft revision needs review and later explicit confirmation.	Destructif
	`revise_study_plan`	Revise Study Plan. Revises one saved draft at its current draft.id and draft.revision without starting research. Accepts edits, refinement, clarification answers or response-format corrections. Stale revisions fail. Fixed questions stay verbatim when only clarification answers change. Bounded history may prune older revisions; the changed draft requires review again.	Destructif
	`get_study_plan`	Read Study Plan. Reads the latest saved revision and full review of one draft. Creates or changes nothing and starts no research.	Lecture seule
	`render_study_plan`	Show Study Plan. Shows a saved Study draft at its current revision for interactive review and confirmation. Research has not started.	Lecture seule
	`read_study_summary`	Read Study Summary. Reads saved whole-Study Markdown, evidence blocks and metadata. Does not regenerate or replace the summary.	Lecture seule
	`refresh_study_summary`	Refresh Study Summary. Generates and saves a whole-Study summary after explicit approval. Replaces the prior summary without restore history; answers and questions are preserved. force regenerates unchanged coverage. Returns Markdown and evidence blocks.	Destructif
Planification de la recherche	`start_study_run`	Start Confirmed Study Run. Runs one reviewed draft revision after explicit confirmation in a later user turn; skipping review is not confirmation. One execution submits the whole draft as a single durable run. It is refused as a whole before using Minds or quota if the revision, capabilities, method or source-readability checks fail. Advanced methods need explicit opt-in. Each Mind answers in order seeing its own earlier answers; askIf gates items. Contradictions are flagged, never rewritten.	Écriture
	`resume_study_run`	Resume Study Run. Resumes an existing plan-limited Study run in place after explicit user approval and the required paid-plan entitlement. Keeps the originally confirmed questions and response contracts and does not ask questions already answered. Does not start a new draft or change the run design.	Écriture
	`list_research_methods`	List Research Methods. Reads the fixed Minds research-method catalog in one operation, including availability, complexity and fallback method. includePlanned only filters the returned catalog. It does not create, configure or execute a method, and the catalog cannot be edited through this tool.	Lecture seule
Modèles et brouillons	`list_study_drafts`	List Study Drafts. Lists a page of unfinished Study draft summaries, newest first, or returns the complete saved planning state for one exact draft ID. Draft records are distinct from running or completed studies.	Lecture seule
	`list_study_templates`	List Study Templates. Lists your own and team-shared Study templates, most used first, as a page of summaries (templateId, name, revision, method, question count), or returns one exact templateId with its revision, research method, questions, response settings and question attachments. configuration.questions is a fixed, pre-registered instrument and configuration.methodId the research method it runs with; both can be planned as they are.	Lecture seule
	`save_study_draft`	Save Study Draft. Creates or checkpoints an unfinished Quick or Custom Study draft without starting research. It saves the objective, context, selected Audiences, method, questions, sources, and current planner step. Revisions change only the fields given and require the exact draft ID and expected revision; stale writes are rejected. For a retryable creation, choose idempotencyKey before the first save and reuse it after an uncertain result.	Destructif
	`save_study_template`	Save Study Template. Creates a private Custom template with its questions, response settings, attachments and method. template.requestId enables safe retries. Does not change existing templates, create a draft or start research.	Écriture
	`update_study_template`	Update Study Template. Updates one owned Custom template at update.expectedRevision. Replaces supplied fields without restore history and may share it with the team. Requires explicit approval; stale revisions fail. Does not start research.	Destructif
	`use_study_template`	Create Draft From Study Template. Creates an independent editable Study draft from a readable template at use.expectedRevision. Reuse use.requestId for retries of the same intended draft. Fresh context and Audiences are added in the web app; the template is unchanged and research does not start.	Écriture
	`delete_study_template`	Delete Study Template. Only for saved Study templates; it cannot delete Studies, Audiences, Minds or answers. Permanently deletes one template owned by the authenticated user. Studies and drafts already created from it are unaffected. Teammates with shared access cannot delete a template they do not own.	Destructif
Minds	`export_mind`	Export Mind Persona Profile. Generates a branded profile for one existing Mind, identified by exact ID or name (partial names and typos resolve to the closest). Markdown is returned inline by default; PDF, DOCX, and PPTX artifacts are returned as base64 with a workspace link.	Écriture
Minds	`get_shared_mind_knowledge`	Read Shared Mind Sources. Read shared Mind sources and assessments.	Lecture seule

Aperçu interactif

Minds enregistre actuellement trois vues MCP : résultats des Studies (pour une nouvelle question et pour une Study existante), revue du plan de recherche et création d’Audiences. Elles utilisent les composants Minds UI ; le client détermine le cadre et les interactions.

La vue du plan affiche le brouillon enregistré. Demandez les modifications dans la conversation, puis confirmez explicitement la dernière révision avant son exécution. Afficher le plan ne lance pas de Study.

Sur un hôte mobile détecté, le widget affiche un lien pour continuer dans Minds. Les commandes interactives et l’actualisation périodique du widget ne démarrent pas. Cela concerne la présentation, pas la disponibilité des outils MCP. Les restrictions du client restent applicables.

Les exemples ci-dessous utilisent des données illustratives. Chaque onglet est une vue, avec les outils qui l’ouvrent ; les états de chargement et en cours sont des instantanés fixes, pas une recherche en cours.

3 widgets · 13 états d’aperçu

Explorer chaque widget MCP

The first render, before the host delivers the question's data.

Données illustratives · l’apparence varie selon le client

Workflows n8n

Utilisez le nœud communautaire Minds pour n8n pour créer des Études, prévisualiser des plans de recherche, récupérer des Études et des résumés enregistrés, ou exposer une opération à un Agent IA. Installez n8n-nodes-minds sur n8n auto-hébergé et connectez une clé API Minds. Le package est publié sur npm ; la vérification n8n est en cours d'examen, il n'est donc pas encore disponible sur n8n Cloud. Examinez et démarrez la recherche séparément dans Minds.