Integrations.
Your tools.
Safely connected.
Connect your tools – sensitive content is anonymized before the AI sees it.
Integrations connect anymize to your external services via OAuth: Microsoft 365, Google Workspace, HubSpot, Notion, Asana, GitHub and more. Once connected, the AI can read and write there – as a tool in chat, as a building block in workflows. The crucial part: data flowing from these apps into the AI can run through anymize anonymization. An Outlook email gets summarized without plain-text PII reaching a US model.
Interactive · Preview
Your apps answer with you.
Connected apps become tools for the AI: you ask in chat — the AI searches Google Drive, files the note in HubSpot and answers with evidence. Same with Microsoft 365, Notion, Asana and any custom connector.
Message the AI …
What are connectors?
Your tools become
tools for the AI.
A connector is an OAuth connection to an external service – your inbox, your CRM, your repository, your database. Once connected, anymize can read and write there on your behalf. You never hand over passwords; you authorize access through the provider's official OAuth flow. Beyond the ready-made connectors, you can also add any remote MCP server yourself by URL.
Connectors take effect in two places – and that is the decisive point:
Two points of use
In chat as a tool
The AI agent calls the connector on its own when your question requires it. "Summarize my last ten Outlook emails" – the AI reads them without you copying anything.
In workflows as a building block
In automated flows, connectors become steps: read data from Google Sheets, create a Notion entry, send an email via SMTP – chained, triggered, repeatable.
The effect: anymize does not stay locked in its own window but works where your data already lives. And everything that comes in can run through anonymization before an AI model sees it.
Available connectors
Eight services, instantly connected.
More on request.
The most important office, CRM, project and developer services connect directly – via OAuth in seconds:
New · Tax & law
TaxGraph — German tax law, right in the chat.
Statutes, Federal Tax Court rulings and Ministry of Finance letters, updated daily and backed by sources — for tax work that must stay confidential and still answer reliably.
How TaxGraph fits anymize →New · Science & research
Elicit — scientific evidence, right in the chat.
Research across 138 million papers, systematic reviews and data extraction, with sentence-level citations — for teams that need reliable evidence and still protect sensitive context.
How Elicit fits anymize →Microsoft 365
Microsoft 365
Outlook (email), Calendar, OneDrive, Teams, Microsoft To Do, OneNote
Google Workspace
Google Workspace
Gmail, Google Calendar, Google Sheets, Google Drive, Google Docs
CRM & projects
HubSpot
Contacts, Deals, Companies, Tickets
Asana
Tasks, Projects
Knowledge & documents
Notion
Create pages (Markdown → Notion blocks)
Developers & data
GitHub
Repositories, Issues, Pull Requests, Code
Postgres
SQL query against an external database
Communication
Email / SMTP
Sending (text + HTML, attachments)
New · Custom connectors
Or connect any MCP server yourself.
You're not limited to our list: add any remote MCP server by URL – with no auth, an API key, or OAuth with automatic discovery. Connect your own or third-party tools without waiting for a built-in integration.
Under control – also across teams
- Per-tool permissions: read and write rights toggled individually
- Always ask on write access – enforceable as a policy
- Group-scoped sharing: only selected teams see and use a connector
- Third-party servers run through the EU privacy subagent; credentials encrypted (AES-256-GCM), with SSRF protection
More categories on request
Need a service that is not yet on the list? The following categories can be connected on request:
How it works
Connect, then use
in chat and in workflows.
Three steps from click to working AI – no passwords, no API keys by hand.
01
The path to a connector
Connect via OAuth
Under "Connectors" you pick the service, get redirected to the provider and authorize access. anymize stores no passwords – only the token issued by the provider.
Use in chat
The AI agent accesses the connector on its own when your question requires it. "Which deals are due this week?" – the AI queries HubSpot and answers.
Use in workflows
In the workflow builder you drop the connector into your flow as a building block: read, transform, write – triggered manually, on a schedule or by an event.
Management
All connections in one place
Under "Connectors" you manage all OAuth connections centrally: connect, review permissions, disconnect. A disconnected connection instantly revokes all AI access to that service.
02 · The security leverage
Content runs through anonymization
What comes out of your apps is anonymized before the AI sees it.
A connector often delivers highly sensitive data – client emails, CRM contacts, patient records. Before this content goes to an AI model, it can run through anymize anonymization.
Practical consequence: an Outlook email with plain-text names, addresses and IBANs gets summarized without the PII ever reaching an international frontier model. The connector fetches the data, anonymization protects it, the model sees only placeholders.
03 · Usage patterns
Three typical connector flows
Inbox in chat
"Summarize my unread Outlook emails"
Outlook connector reads the mails – anonymized before the model
Compact summary, no plain-text PII at the US model
CRM in a workflow
Weekly workflow starts automatically
HubSpot connector pulls due deals, the AI prioritizes
Triage list lands as an email via SMTP in the team inbox
Report from data
"Build the monthly report from the revenue sheet"
Sheets connector reads numbers, Postgres adds a detail query
Finished report as an artifact, optionally saved to Notion
Anonymization in integrations
Data from your apps
runs through anonymization.
This is the core difference to any other connector system: content flowing from Outlook, HubSpot, Drive or Postgres into the AI does not go to an international frontier model in plain text. It passes through anymize anonymization first. The model sees placeholders – yet you still get the answer about your real data.
The path of a record
An Outlook email with plain-text names, addresses and IBANs is fetched by the connector, anonymized and only then handed to the model. The answer is translated back automatically – you read "Dr. Maria Schmidt", the model only ever saw [[LASTNAME-7F2]].
What this means
Sensitive content stays protected
Client emails, CRM contacts, patient records, database fields – everything a connector delivers can be anonymized before it reaches a model.
Full answer quality
Bidirectional anonymization translates the answer back automatically. You keep working with your real data – the model never saw it in plain text.
Connect your tools – sensitive content is anonymized before the AI sees it. This holds in chat and in workflows, equally for every connector.
Use cases
What connectors
actually deliver.
Five flows born out of real work:
Outlook · Gmail
Chat
Mails anonymized before the model – no plain-text PII at the US model
HubSpot
Chat · Workflow
Query contacts, deals and tickets, let the AI prioritize them
Google Sheets · Postgres
Workflow
Read numbers, generate the report, optionally export as an artifact
HubSpot · Asana
Workflow
Classify incoming tickets, assign, set status
Google Drive · OneDrive
Chat
Read and summarize files, sensitive content anonymized
The pattern
Connectors live where your data already is. You no longer fetch it into the AI by hand – the AI fetches it itself, anonymizes it and gives you the result. On request in chat, automatically in workflows.
What you should know about integrations.
Frequently asked questions
An OAuth connection to an external service – such as Microsoft 365, Google Workspace, HubSpot or GitHub. Once connected, anymize can read and write there on your behalf. Connectors take effect in two places: in chat as a tool the AI agent calls on its own, and in workflows as a building block for automated flows.
We stand behind anymize. And we know – when an AI tool touches client, patient or employee data, a demo video isn't enough. That's why we give you 14 days of full access – all models, all features, no credit card. Enough time to be certain, before you trust us.
Your AI workplace awaits.