Subprocessors

Last updated: August 10, 2026

These are the third-party services that may process customer data on Tidela's behalf. We keep this list short on purpose — every name here is a company your data can touch, and a dependency we have to stand behind. If you want notice when it changes, email security@tidela.ai and we'll add you.

Core infrastructure

These process customer data as a matter of course, for every customer.

• DigitalOcean — application hosting, database, and file storage. United States. This is where your Tidela workspace lives.
• Anthropic — the Claude models behind AI Selling, and the default AI provider. United States. Receives the specific record content needed to answer the request a user just made. Used under Anthropic's commercial API terms; Tidela does not permit customer data to be used for model training.
• Postmark — delivery of system email such as sign-in links and notifications, and mail your team sends from Tidela. United States.

If you bring your own AI key. A workspace can supply its own key for another AI provider — OpenAI, Google, or an OpenAI-compatible endpoint among them. If you do, prompts containing your record content go to the provider you chose instead of Anthropic, under your agreement with them rather than ours. This is off unless you turn it on.

Operational services

Used to keep Tidela running and safe. These see limited data — error diagnostics, or a single address at a time — not your CRM records in bulk.

• Sentry — application error monitoring. Receives technical diagnostic data when something breaks; personal data is not attached by default.
• Cloudflare (Turnstile) — bot protection on sign-up and sign-in-link requests.
• OpenStreetMap — two uses. Nominatim converts a street address into map coordinates, receiving that address and nothing else; the tile servers render the map itself, which means they see the browser's IP address and the area of the map being viewed.
• Company logo and avatar lookups — to display a logo next to an account, Tidela requests it from Clearbit, Google's favicon service, or DuckDuckGo's icon service, sending the account's website domain. Contact avatars are looked up at Gravatar using a one-way hash of the contact's email address. These are cosmetic; tell us and we'll disable them for your workspace.
• Content delivery networks — Google Fonts serves the typeface, and unpkg serves the mapping and org-chart libraries used on those two screens. They see the browser's IP address, not your records.
• Browser push services (Apple, Google, Mozilla) — if a user enables push notifications, their browser's push service delivers the alert.

Optional, only if you connect them

Nothing here is active unless a user in your workspace explicitly authorises it, and you can revoke that authorisation at any time.

• Google (Gmail & Calendar) — mail and calendar sync, if a user connects their Google account.
• Microsoft (Microsoft 365 / Graph) — mail and calendar sync, if a user connects their Microsoft account.
• ElevenLabs — speech narration for the in-app Training walkthroughs. Receives Tidela's own help and training script text, not your customer data.
• Your previous CRM — if you use a migration connector to import from HubSpot, Pipedrive, Attio, Nimble, Copper, Capsule, monday.com, or Pipeliner, you give Tidela credentials for that system and we read your records out of it. Data flows in, not out.

One thing that isn't a subprocessor, but you should know

Tidela runs its own business on Tidela. When you start a trial, your company and the users on the account are created as records in our internal workspace, so we can support you and see how the trial is going — the same way any vendor keeps you in their CRM. It stays inside Tidela, it is not shared, and it is deleted when your workspace is deleted.

How we manage them

We pick vendors that publish their own security posture, we send them the minimum data needed for the job, and we hold their credentials encrypted. Before adding a subprocessor that would handle customer records, we review what it receives and why — and it shows up on this page. This list is maintained by hand, is verified against the code rather than from memory, and reflects what is actually wired into the product today.