Encryption, isolation, authentication, and governance built into the core — so your autonomous workforce is auditable, gated, and yours to control. This page is written to survive a security review.
Multi-tenant isolation, four levels deep.
A strict hierarchy segregates every byte of data: App Admin → Partner → Tenant → User. Every database operation is company-scoped — no query can cross a tenant boundary. Full data segregation is the default, not a configuration.
It asks permission at every checkpoint you define, runs every model choice on the record, and stops the moment a budget is breached.
AES-256-GCM with nonce-prepended ciphertext. The master key is stored outside the database. Secrets are never logged or exposed in plaintext.
JWT access tokens with 15-minute expiry plus HTTP-only refresh cookies (7-day). A suspension middleware instantly locks out suspended tenants at the API.
Configurable checkpoints per agent, action, process, and loop. Pause before large payments, contract signatures, or promoting self-evolving code — surfaced to you through Pragya.
Restrict which models are eligible (residency/compliance) or pin a specific one. Every step records the model, tokens, and cost. Routing is completely auditable.
Minimum wallet thresholds per execution type and real-time per-step cost tracking. Execution halts automatically on a budget breach. No runaway spend.
Every run traced: token counts, cost attribution, latency, and decision logs. Exportable, queryable, compliance-ready — with context snapshots for replay.
You decide which brains touch your data — and you can prove what every one of them did. Talk to us about your security requirements.