Security at TODOforAI
TODOforAI logs into the services your company already uses and gets your todos done. That only works if you can trust us with access — so security is the core of the product, not an afterthought. Here is how we protect your data, your credentials, and your accounts.
EU Infrastructure
- Customer data is hosted in the EU (Hetzner, Finland/Germany), operated by an EU company (TODO for AI Kft., Budapest, Hungary).
- Cloudflare provides network security, DDoS protection, and TLS termination at the edge.
- Backups are encrypted and stored within the EU.
Your Credentials. Your Control.
Connected tools handle authentication within your workspace. In the normal connected-tool flow, agents receive task results, not your login credentials. Connecting an account is not blanket permission to act: access is checked per task.
- Secrets stored in the TODOforAI vault are encrypted at rest. Connected tools may also retain login tokens in their own credential stores within your workspace so they can stay signed in between tasks.
- This describes normal authentication, not a guarantee that credentials can never appear in tool output or be accessed by software with permission to read the credential store.
- Connect only the accounts you choose. Revoke access with your provider at any time.
Our Workspace Access Policy
Our staff do not routinely access your private workspace or inspect its contents, including your connected credentials.
We administer the underlying cloud infrastructure. Our policy limits exceptional administrative access to security, maintenance or user-requested support purposes. This is an access policy, not a claim that administrative access is technically impossible.
Running tasks necessarily involves processing your data. Task content and tool results may be sent to the AI providers used for your task. Keeping login credentials out of normal tool responses does not mean task data stays entirely on your computer. See our GDPR & Data Processing page for more information.
Isolated Workspaces
- Cloud workspaces use virtual-machine isolation to separate customers' workloads.
- Workspace files and connected-tool login state can persist between tasks. Stopping an agent or a virtual machine does not by itself erase that data or revoke connected accounts.
Encryption Everywhere
- TLS 1.2+ for all external connections.
- Agent/edge communication uses the Noise Protocol Framework for end-to-end encrypted transport.
- Data encrypted at rest, including backups.
Human-in-the-Loop by Design
The agent never auto-runs irreversible or money-touching actions. Charging, cancelling, or changing billing is always prepared by the AI and approved by a human with one click. Low-risk work runs autonomously; the big calls stay yours.
Audit Logs
Every tool call the agent makes is logged — what was done, when, and in which service — so you can always answer "what did the AI actually do?"
Your Data Is Not Training Data
- We don't train AI models on your business data.
- AI providers (Anthropic, OpenAI, Google) are used via their API terms, under which your data is not used to train their models.
- We process only what a todo needs — data minimisation by default.
GDPR & Compliance
- Full GDPR compliance as an EU company — see GDPR & Data Processing for roles and our sub-processor list.
- Data Processing Agreement (DPA) available for signature — covers breach notification within 72 hours, data deletion, and audit rights.
- Data transfers outside the EU are protected by the EU–US Data Privacy Framework or Standard Contractual Clauses.
- We support enterprise security reviews and vendor questionnaires — contact us and we'll work through yours.
Open Source
Our edge agent and CLI are open source — you can inspect exactly what runs on your machines.
Reporting a Vulnerability
Found a security issue? Email security@todofor.ai. We respond promptly and appreciate responsible disclosure.
For security reviews, DPAs, or any other questions: privacy@todofor.ai