MCPFeature · AI assistant · open standard

Let your AI assistant work inside your CRM.

vemix ships an MCP server. You connect it once via OAuth, with Claude, ChatGPT or any other MCP client. Your assistant then researches, drafts emails and creates companies, contacts and tasks directly in vemix. Leads, documents and presentations land there too.

Start your free trial Open vemix

Works with Claude ChatGPT+ more MCP clients

OAuth 2.1 / PKCE role- & workspace-bound every call permission-checked

ClaudeMCP client · example

„Research Atlas Cloud GmbH, create the company & contact in the CRM and prepare a first email."

vemix.createCompany()▍

vemix · CRMwritten live

Atlas Cloud GmbH

Company · Munich

Dr. Lena Behrens

Contact · CTO

Lead · First contact

Pipeline · New

MCP · OAuth

Workflow 1

Research, create, prepare

Your AI assistant researches a prospect, writes the result straight into vemix and prepares the next step so your team can start right away.

one prompt → four steps

Research

Market, people and context on the prospect, gathered by the assistant.

Research note

Create in the CRM

Company, contacts and a lead in the right pipeline, written via MCP.

Company

Lead

Prepare the task

A follow-up task with a ready-to-send email draft on the contact.

First email · draft

due tomorrow

Generate the deck

A tailored presentation or document, filed in vemix.

.docx

.pptx

Workflow 2

Draft, send and file automatically

Prepare the documents in your AI assistant, send them from your own mailbox and put your vemix archive address in BCC. vemix creates the contact and company and links the email automatically.

Draft in the assistant

Offer, memo or letter in the chat.

Send via M365 / Google

From Outlook or Gmail, with the CRM in BCC.

Recipient

Receives the mail from your mailbox.

BCC · archive

Auto-filed

vemix creates contact & company and links the email.

What the AI can do in vemix

The MCP toolset

Every action runs in your name, is bound to one workspace and respects your roles and permissions. The assistant only sees and does what you are allowed to.

Companies & contacts

Read + write

Create and update companies, contacts, emails and phone numbers straight from the chat.

Leads & pipelines

Read + write

Read pipelines, create leads in the right stage and move them as the deal progresses.

Tasks with drafts

Read + write

Create follow-up tasks and prepare the matching email draft so the work is ready to go.

Documents & decks

Generate

Generate Word documents and PowerPoint presentations and file them on the record.

Contracts & signatures

requires the module

Draft contracts and kick off signature requests from the chat.

Role- & workspace-bound

Every tool respects roles, ownership and module access and is bound to one workspace.

How access works

Connect once, your permissions always apply

The MCP server uses OAuth 2.1 / PKCE, and you choose the workspace when you grant consent. Every tool call is permission-checked. Remove a user and their access ends immediately.

AI assistant

connects once

OAuth 2.1 / PKCE

workspace choice on consent

vemix MCP server

app.vemix.io/api/mcp

Part of API & MCP access every call permission-checked removal ends access instantly

Unattended automation?

For server-to-server use (for example an n8n chatbot that creates a contact on a callback request) vemix offers a scoped public API with self-managed keys, separate from the interactive MCP connector.

FAQ

Frequently asked questions

Which AI assistants work with vemix?

Any MCP-capable client. Claude and ChatGPT are among the tested ones; through the open MCP standard you can connect further assistants and agent frameworks just as well. The actions are the same; only the client differs.

How does an AI assistant connect to vemix?

Through the MCP server at app.vemix.io/api/mcp. You connect it once via OAuth 2.1 / PKCE and choose the workspace; from then on the assistant works in your name with your data. This requires the API & MCP access included in your edition.

What happens to my data?

Your assistant retrieves data from vemix and processes it at the provider of your AI assistant, for example Anthropic (Claude) or OpenAI (ChatGPT). This runs through your own account with that provider, under its terms and, depending on the provider, possibly outside the EU. That provider is not a sub-processor of vemix: choosing it, the contract or data processing agreement with it, and the legal basis are the responsibility of your company as the vemix customer. The assistant only gets access after your OAuth consent, every call is permission-checked, and you can disconnect it at any time in vemix under Settings → MCP server (see also "How access works" above). Removing the connector in the assistant alone is not enough to revoke access. Data covered by professional secrecy under section 203 of the German Criminal Code (StGB), such as patient data or client data of lawyers and tax advisers, does not belong in vemix and must not be retrieved through the AI assistant either.

Can the AI only read, or also create data?

Both, within your permissions. A viewer sees only read tools; with create rights the assistant can add companies, contacts, leads, tasks, documents and more. Every tool call is permission-checked, exactly like the app.

How do email drafts and presentations reach you?

The assistant creates the task, writes the prepared email draft onto the contact and generates documents or PowerPoint decks that are filed on the record. You just review, adjust and send.

What about unattended automation?

Put your AI assistant to work in vemix.

Start your free trial and connect the MCP server via API & MCP access.

Start your free trial Open vemix