Security
Security built for your company.
Canu connects to important business systems, so security is built into every layer of the product — from authentication and encrypted connections to permissions, approvals, and audit history.
Overview
Your data stays protected.
Canu only accesses the information needed to perform the tasks you authorize. Connections, credentials, agent actions, and company data are protected using modern security practices.
Encryption
Encrypted in transit and at rest
Data transmitted between Canu and connected services is protected using encrypted HTTPS/TLS connections. Sensitive stored data is also encrypted at rest using secure managed infrastructure.
Secure connections
Credentials stay protected
Canu uses OAuth and secure credential-management systems wherever integrations support them. You never need to hand your account passwords to Canu.
Permission controls
Only the access Canu needs
Connections follow least-privilege principles, so Canu receives only the permissions required for the actions your company enables.
Approval controls
You stay in control
Sensitive or higher-impact actions can require explicit human approval before Canu executes them.
Connections
Secure connections to your tools
Canu connects through secure APIs and OAuth flows wherever they are available. Raw account passwords are never requested or stored.
Canu connects through secure APIs and OAuth flows whenever possible.
You authorize
You choose which company systems Canu can connect to.
Canu receives limited access
Canu only receives the permissions required for the tasks you enable.
You can disconnect anytime
Connections can be revoked from the Canu Connections page.
Credentials
Canu does not need your passwords.
Whenever an integration supports OAuth or token-based authentication, Canu uses that secure authorization flow instead of asking for account passwords.
Access tokens and integration credentials are stored using secure encrypted credential infrastructure and are never exposed in the frontend or client-side code.
Agent permissions
An AI agent should never have unlimited access.
Canu's capabilities are grouped into permission tiers. Your company decides which tiers Canu is allowed to use, per integration.
Read
- View expenses
- Read usage
- Analyze contracts
- Review subscription activity
Prepare
- Prepare cancellation
- Draft vendor outreach
- Create recommendations
- Generate cost-saving actions
Execute
- Cancel a subscription
- Change a service configuration
- Send an approved message
- Modify a vendor setting
Execution permissions are separate from read permissions.
Granting Canu the ability to read your spend never implies the ability to change it.
Approvals
Important actions can require your approval.
Remove 12 inactive Slack seats
- Expected savings
- $144/mo
- Risk
- Low
- Affected users
- 12
Companies configure which types of actions Canu may execute automatically and which must be approved by a human first.
High-impact actions default to requiring approval, and approval checks are enforced on the backend — not in the browser.
Agent safety
Built for agent safety
Canu does not let an AI model perform arbitrary actions. The agent operates through a controlled tool layer where every action passes permission and approval checks.
Restricted tools
The agent can only call integrations and actions explicitly exposed to it.
Permission checks
Before execution, the system verifies whether the workspace allows that action.
Approval gates
Actions requiring human approval pause until a person approves them.
Verification
After execution, Canu checks whether the expected result actually occurred.
Data privacy
Your company data is your company data.
Canu only uses connected company information to provide the Canu product and perform the agent tasks you authorize.
- Company data is not publicly exposed.
- Access is scoped to the appropriate workspace, enforced server-side.
- You can remove integrations and revoke access at any time.
- Sensitive credentials never appear in application logs or frontend interfaces.
AI providers
How AI is used
Canu uses AI models to understand company information, analyze financial and operational data, identify opportunities, and decide which authorized tools to use. The model never receives unrestricted access to your infrastructure — tool access is mediated by Canu's backend permission system.
Infrastructure
Secure infrastructure
Backend-only secrets
API keys and integration secrets stay server-side and are never shipped to the browser.
Environment isolation
Production credentials are stored separately from development environments.
Encrypted communication
All application and API traffic uses HTTPS/TLS.
Database protections
Encrypted storage, authenticated database access, and strict server-side authorization on every request.
Logging
Important security and agent events are recorded without logging sensitive credentials.
Backups
Secure backups for critical company data where applicable.
Audit history
See what Canu did.
Every meaningful agent action leaves a visible audit trail — the user, the action, the time, the integration, its status, the approval source, and the verified result.
Account
Protecting your Canu account
Access to the Canu workspace itself is protected with standard account security controls.
Workspace access
Coming SoonWorkspace-level access controls
Multiple employees will be able to use Canu without identical permissions. These roles are not yet available in the product.
Admin
Can manage connections, permissions, and approvals.
Approver
Can review and approve agent actions.
Member
Can view analysis and interact with Canu.
Viewer
Read-only access.
Disconnect anytime.
Workspace admins can revoke an integration directly from Canu's Connections page. When disconnected:
- Canu stops making new requests to that integration.
- Stored access tokens are revoked or securely removed.
- Tasks depending on that integration stop.
- The activity log records the disconnection.
Architecture
How a Canu action actually flows
Every action travels the same path — from an authorized connection, through permissions and approvals, to a verified, logged result.
Connected services
QuickBooks / Stripe / AWS / Slack / OpenAI
Secure OAuth + API layer
Encrypted credential store
Canu backend
Permissions engine
AI agent
Approval engine
Authorized action
Audit log + verification
Security incidents
If Canu identifies suspicious access, compromised credentials, or another security issue, affected access is disabled and investigated as quickly as possible. You always have a clear way to report a security concern.
Found a security issue?
We appreciate responsible disclosure from security researchers and users. Please contact our security team with details about the issue and steps to reproduce it.
Compliance
Security and compliance
As Canu grows, we plan to continue formalizing our security program and pursue industry-standard security reviews and certifications appropriate for our customers. The items below are not yet in place.
SOC 2
Penetration testing
SSO
MFA
RBAC
FAQ
Security questions, answered
Connect your company with confidence.
Canu is designed so your AI agent gets the access it needs — and nothing more.