# Let your AI run your business > Ondes PRO’s MCP server connects Claude, Cursor, ChatGPT or VS Code to your store, invoices and projects. OAuth, test mode, included in every plan. Updated on 2026-10-21 · The Ondes PRO team · https://ondes.pro/en/integrations/mcp Every Ondes PRO workspace ships with an **MCP server** (Model Context Protocol). Your AI assistant — Claude, ChatGPT, Cursor, Visual Studio Code… — connects in a minute to your store, inventory, quotes, projects and calendar. It reads real data, acts with your permissions, and changes nothing without your approval. ## One sentence, several modules _Example conversation · demo data (Atelier Lumen)_ > **You** : Hôtel Ker Aven wants 40 Pétrole pendant lights by 15 November. Draft the quote, check stock and add the delivery to the schedule. > > 🔧 `ondes_appeler` `GET inventory/composants` — 26 Pétrole pendants in stock, 30 being manufactured. > > 🔧 `ondes_appeler` `POST billing/devis` with `simuler: true` — the quote is valid, nothing written yet. > > **Assistant** : Stock covers 26 units; the running production order delivers the other 14 on 8 November. I suggest a quote of **40 × €189 excl. VAT**, a “Ker Aven delivery” card on 15 November in the Logistics board and an installation slot on the 14th. Shall I create all three? > > **You** : Yes, go ahead. > > 🔧 `ondes_appeler` Quote DEV-2026-0142 created, Kanban card and calendar event added. > > **Assistant** : Done. The quote is a draft in Invoicing, so you can review it before sending. ## What is an MCP server, in two sentences? The **Model Context Protocol** is an open standard, launched by Anthropic in 2024 and since handed to the Linux Foundation’s Agentic AI Foundation, that lets an AI assistant plug into software the way a USB stick plugs into a computer. The software exposes tools; the assistant discovers them and uses them when your request calls for it. In practice, you no longer export a spreadsheet to paste it into a chat: the assistant fetches up-to-date information itself and can act inside your business software. To dig deeper, read our [complete guide to the Model Context Protocol](https://ondes.pro/en/integrations/mcp/what-is-mcp). - **1 min** — to connect an assistant, with nothing to install server-side - **8 tools** — stable, giving access to the API’s ~135 resources - **2,700** — tokens of permanent context, instead of 44,000 for a documentation file - **1 h** — lifetime of an access token, revocable at any time ## Connect your assistant - [Claude](https://ondes.pro/en/integrations/mcp/clients/claude) — Claude Desktop, claude.ai and Claude Code: one custom connector, one OAuth sign-in, and you are set. - [ChatGPT](https://ondes.pro/en/integrations/mcp/clients/chatgpt) — ChatGPT developer mode accepts your workspace as an MCP connector, over OAuth. - [Cursor](https://ondes.pro/en/integrations/mcp/clients/cursor) — One `.cursor/mcp.json` file, one OAuth sign-in: Cursor’s agent codes against your real workspace. - [Visual Studio Code](https://ondes.pro/en/integrations/mcp/clients/visual-studio-code) — GitHub Copilot agent mode reads and calls your workspace from `.vscode/mcp.json`. ## The shortest path: Claude Code One command, then `/mcp` inside Claude Code: a window from your workspace opens, you sign in, and the assistant can reach your modules. Where to set it up : One terminal command, then `/mcp` inside Claude Code to sign in. **OAuth connection (recommended)** (Terminal) ```bash claude mcp add --transport http ondes https://your-workspace.ondes.pro/mcp/ ``` **API key connection** (Terminal) ```bash claude mcp add --transport http ondes https://your-workspace.ondes.pro/mcp/ \ --header "Authorization: ApiKey opk_…" ``` _Tested by our team. Vendor documentation : https://code.claude.com/docs/en/mcp_ ## What your AI can do in Ondes PRO - **Sell**: find an order, create a promo code, update a product, track a shipment — see [the online store](https://ondes.pro/en/integrations/mcp/modules/online-store). - **Invoice**: draft a quote, turn it into an invoice, list overdue invoices, issue a credit note — see [invoicing](https://ondes.pro/en/integrations/mcp/modules/invoicing). - **Manage stock**: check levels, record a delivery, start a production order — see [inventory management](https://ondes.pro/en/integrations/mcp/modules/inventory). - **Follow customers**: enrich a CRM record, add a note after a call — see [the CRM](https://ondes.pro/en/integrations/mcp/modules/crm). - **Organise**: create Kanban cards, schedule a meeting, set a recurring appointment — see [Kanban](https://ondes.pro/en/integrations/mcp/modules/kanban) and [the calendar](https://ondes.pro/en/integrations/mcp/modules/calendar). - **Communicate**: sort and answer emails, prepare a campaign — see [email](https://ondes.pro/en/integrations/mcp/modules/email). - **Automate**: describe a flow in one sentence and get a draft automation — see [automations](https://ondes.pro/en/integrations/mcp/modules/automations). ## Eight tools, and not one more Exposing every resource as a tool would flood the assistant’s memory. The server offers eight stable tools: six to understand the API without guessing, two to act. - `ondes_etat` (Read) — **Connection status** : Who is connected, which modules are reachable and which permissions the connection carries. The assistant calls it once to learn its scope. - `ondes_guide` (Read) — **API guide** : The API guide, section by section: a summarised table of contents first, then only the sections that matter (pagination, idempotency, payments…). - `ondes_chercher` (Read) — **Search** : Searches a business term (“quote”, “stock”, “webhook”) across both the resource inventory and the guide. - `ondes_ressource` (Read) — **Resource sheet** : Everything needed to call one resource correctly: URL, verbs, typed and required fields, filters, ordering, actions, required permission, test twin. - `ondes_impossible` (Read) — **What the API does not do** : The list of what deliberately does not exist. The assistant checks it before inventing a URL: this is what keeps it from hallucinating. - `ondes_exemple` (Read) — **Complete recipes** : Whole scenarios already verified against the real API: create, read without N+1 calls, invoice, subscribe to events. - `ondes_appeler` (Write) — **Call the API** : Runs a real call with the user's identity. Writes go to test mode by default, and a dry run validates a request without writing anything. - `ondes_assistant` (Write) — **Ask the Ondes PRO assistant** : Asks a business question to the built-in assistant, which aggregates several modules. Its writes come as a plan, executed only after approval. ## Documentation that never goes stale Until now, plugging an AI into software meant handing it a large documentation file. Ours weighed **44,000 tokens**: a quarter of an assistant’s working memory, for a document it only uses a few percent of — and which went out of date the moment it was copied into a repository. The MCP server serves the same documentation **on demand**, section by section, generated on every call from your workspace’s real routes. It only describes the modules you can access, and the `ondes_impossible` tool lists what the API deliberately does not do: that is what stops the assistant from inventing a URL that would return 404 for your customers. ## Your AI breaks nothing by accident Letting an AI write to your data takes safeguards. They live in the server, not in the model’s good intentions. - **Test mode by default**: when a resource has a test twin (orders, meetings, email sends, invitations), writes go there unless the assistant explicitly asks for the real thing. - **Dry run**: `simuler: true` validates a write and shows what would be created, without saving anything. - **Two-step confirmation**: Ondes PRO’s built-in assistant proposes its changes as a plan, executed only after you approve. - **Idempotency**: a request replayed after a network drop never creates a duplicate. - **The mode is always stated**: every response says whether the write was a test or real. > **An MCP connection acts with your permissions — all of them** — It is a deliberate choice, and we would rather say it plainly: a connection authorised through OAuth carries your full identity, no more and no less. In return, the access token lives **only one hour**, every connection is **named per machine** (“Cursor, on my laptop”) and can be **revoked instantly**, and every call is logged. For narrower access, use an API key with limited scopes. Everything is detailed on the [MCP server security](https://ondes.pro/en/integrations/mcp/security) page. ![Ondes PRO consent screen (in French): destination, callback address, name declared by the application, account, and a warning about the connection’s full permissions.](https://ondes.pro/integrations/mcp-consentement-oauth.webp) ![List of active MCP connections (in French): Claude Desktop, Claude Code on a MacBook, Cursor on a workshop computer, each with its date, last activity and a Revoke button.](https://ondes.pro/integrations/mcp-connexions-actives.webp) ## By industry - [Ecommerce](https://ondes.pro/en/integrations/mcp/industries/ecommerce) — Orders, stock, promotions and customer service: your AI assistant runs the store alongside you. - [Web agencies and IT services](https://ondes.pro/en/integrations/mcp/industries/web-agencies) — One MCP server per client, documentation that is always accurate, and a test mode to ship without breakage. - [Small businesses and trades](https://ondes.pro/en/integrations/mcp/industries/small-businesses) — Quotes, invoices, reminders and scheduling prepared by AI and approved by you, even between jobs. - [Manufacturers](https://ondes.pro/en/integrations/mcp/industries/manufacturing) — Components, bills of materials, production orders and low stock: the AI spots bottlenecks before you do. ## For developers - [Connect your website](https://ondes.pro/en/integrations/mcp/guides/connect-your-website) — Show the catalogue and take Stripe payments from a site that is already live, without rebuilding it. - [Build an integration](https://ondes.pro/en/integrations/mcp/guides/build-an-integration) — Scoped keys, signed webhooks, idempotency and test mode for a back office that never creates duplicates. - [MCP vs REST API](https://ondes.pro/en/integrations/mcp/mcp-vs-rest-api) — What really sets them apart, when to pick which, and why at Ondes PRO they work together. - [MCP vs Zapier, Make, n8n](https://ondes.pro/en/integrations/mcp/mcp-vs-zapier-make-n8n) — A dated, sourced comparison of two different takes on connected AI, and why they complement each other. ## Frequently asked questions ### Does the MCP server cost extra? No, it is included in every Ondes PRO plan — Essentiel, Pro and Entreprise — with no add-on to enable. See [pricing](https://ondes.pro/en/pricing). ### Which AI assistants can connect? Any MCP client that speaks HTTP: Claude (Desktop, claude.ai and Claude Code), ChatGPT in developer mode, Cursor, Visual Studio Code with GitHub Copilot, Windsurf, Zed, Cline, JetBrains IDEs… Our team has tested Claude end to end; the others follow the standard protocol. ### Do I need to be a developer? No. With Claude or ChatGPT, you paste your workspace address into the connector settings and sign in. Code editors (Cursor, VS Code…) need a configuration file, which each client page provides ready to copy. ### Is my data used to train models? Ondes PRO sends nothing on its own: your assistant reads what it needs to answer. What happens to the conversation depends on the assistant you choose and its privacy settings — check them with its vendor. ### Can I limit what the AI is allowed to do? Yes. An OAuth connection carries your permissions; to restrict it, create an API key with precise scopes (for example read-only invoicing) and use it instead. A key can never exceed its holder’s permissions. ### What if I run several companies? The server respects your workspace partitioning: the assistant targets the company you choose, and a subsidiary never sees a sister company’s data.