Defense-in-depth security across every layer — from authentication to AI safety. Your data stays yours.
Your data is encrypted everywhere it lives and everywhere it moves. You retain full ownership and can delete it at any time.
All sensitive data — API credentials, connection strings, and secrets — is encrypted using AES-256 before it reaches the database. Encryption keys are managed separately from application data, following defense-in-depth principles.
Every connection uses TLS with HTTP Strict Transport Security (HSTS) enforced. Certificates are automatically managed and rotated. All internal service communication stays within private networks, never exposed to the public internet.
Customer data is never used to train AI models. You retain full ownership of all content, conversations, and documents. Request a complete data export or permanent deletion at any time — no lock-in, no hidden copies.
Every organization operates in its own secure boundary. Data isolation is enforced at every layer of the stack — from the database to the storage layer.
Every database query is automatically scoped to your organization. Users, agents, workflows, documents, and conversations all live within your organization's boundary — there is no way to access another tenant's data.
Files and documents are stored in organization-specific paths. Each tenant gets a completely separate storage namespace, preventing any cross-contamination of uploaded files, workflow outputs, or knowledge base documents.
Database-level isolation ensures that even if application logic were bypassed, cross-tenant data access is structurally impossible. Organization context is embedded in authentication tokens and validated on every request.
Multiple layers of identity verification and permission enforcement protect every action on the platform.
Passwords are hashed using industry-standard algorithms. Access tokens are short-lived and automatically rotated via secure refresh flows. Sessions expire predictably and cannot be reused after logout.
Define roles with granular permissions scoped to specific resources and actions. Control who can create agents, execute workflows, manage billing, or administer the organization. Permissions are enforced on every API call.
API keys are cryptographically hashed before storage — plaintext keys are never persisted. Keys support scoped permissions and can be revoked instantly. All API key usage is logged for audit purposes.
Purpose-built protections prevent prompt injection, data leakage, and knowledge base contamination across every AI interaction.
All user inputs are analyzed for instruction override attempts, jailbreak patterns, and context manipulation before reaching the AI model. Suspicious inputs are flagged and blocked, protecting your agents from adversarial prompts.
AI responses are scanned to prevent accidental leakage of system prompts, internal configuration, or sensitive data. Output filters catch and redact information that should never appear in user-facing responses.
Documents uploaded to knowledge bases are scanned and sanitized before processing. Malicious content — including embedded injection attempts — is stripped during ingestion, ensuring your knowledge base stays clean and trustworthy.
RAG search results are strictly scoped to your organization's documents. Agents can only retrieve and cite content from knowledge bases they are explicitly authorized to access — no cross-tenant data leakage is possible.
Production infrastructure follows zero-trust principles with strict network isolation, security headers, and origin validation.
All internal services run on private networks with no direct external access. A hardened reverse proxy is the only externally reachable component, enforcing TLS termination, request filtering, and access policies at the edge.
Every response includes HSTS, Content-Security-Policy, Referrer-Policy, X-Content-Type-Options, and Permissions-Policy headers — enforced at the reverse proxy level. These headers protect against clickjacking, MIME sniffing, and cross-site attacks.
Cross-origin requests are validated against strict allowlists. Only explicitly approved domains can interact with the API. Widget endpoints use controlled CORS policies to enable embedding while preventing unauthorized access.
Embeddable widgets include multiple independent security layers to prevent abuse, even when deployed on third-party websites.
Independent rate limits are enforced per widget, per visitor, and per daily budget. Multiple overlapping controls ensure that no single bypass can exhaust your resources. Usage caps are configurable per widget deployment.
Widgets only respond to requests from approved domains. All user inputs are validated for length and content before processing. Combined with AI safety filters, this creates a robust defense against both automated and manual abuse.
Workflow code steps run in isolated sandboxes with strict resource limits and pre-execution validation.
User-provided code runs in an isolated execution environment with no access to the host filesystem, network internals, or system processes. Each execution gets strict memory and time limits, preventing resource exhaustion.
Code is analyzed before execution to detect and block dangerous operations — including attempts to access the filesystem, spawn processes, or execute arbitrary imports. Only a curated set of safe APIs is available to running code.
Comprehensive logging, infrastructure monitoring, and compliance-ready architecture support your organization's regulatory requirements.
Every security-relevant action is logged with user identity, timestamp, IP address, and request context. Audit logs are immutable and queryable, supporting incident investigation and compliance reporting.
Real-time metrics collection and alerting track system health, request patterns, and queue performance. Anomalous behavior is detected and surfaced through pre-configured dashboards and alert rules.
Every workflow run, agent conversation, and API call is recorded with full input/output details and token usage. Review what happened, when, and why — across every automated process in your organization.
Orckai is built to meet the security and compliance needs of regulated industries, with the flexibility to deploy wherever your policies require.
Architecture designed around SOC 2 trust service criteria — security, availability, and confidentiality controls built in from day one.
Data minimization, right to erasure, data portability, and transparent processing. Built for organizations operating under EU data protection regulations.
Complete data separation between organizations at every layer — database, storage, and application. No shared state, no shared access.
AES-256 encryption at rest and TLS encryption in transit. Sensitive credentials are never stored in plaintext.
Deploy on your own infrastructure for complete data sovereignty. Same features, same security — running entirely within your network boundary.