Not a wrapper around an LLM.
An operating system for AI work.
BBB's Brain is a runtime: a multi-tenant Postgres core, sixty-seven named agents with calibration histories, per-tenant encryption that you control the keys for, and a write-once audit log that every single state change passes through. This is the part competitors usually don't have — because they're not building it.
Multi-tenant by construction · Per-tenant DEK · Write-once audit · 67 agents online
The platform, in pieces.
The Brain is one runtime, but the surface you interact with is split across a few pages — so you can dive into the bit you care about without reading everything else first.
-
Product overview
The full tour of what BBB is. Departments, agents, runtime, audit, governance — in one walk-through.
Read more →
-
Brain Command Center
LiveOne screen for the entire AI workforce. Live ops, agent activity, department health, ROI.
Read more →
-
Skill Tree
A living capability graph for every agent. Mastered, learning, locked, failed — calibrated daily.
Read more →
-
Departments
Fifteen specialist teams. Activate the ones you need, when you need them. Pay for what you turn on.
Read more →
-
Governance & safety
Per-tenant DEK, RLS on every table, write-once audit, approval gate by default. Engineered for procurement.
Read more →
-
Integrations
Six social platforms, Stripe, Supabase Auth, webhooks. Connects to what you already run.
Read more →
Top to bottom, end to end.
Each layer is independently auditable. Each communicates only through declared interfaces. The Brain Core is the foundation; the Customer/Tools surface is what your customers see. Everything in between is the work happening on your behalf.
-
Layer 01
Customers & Tools
Where your business meets the outside world. Six social platforms, BB ERP, Slack/email integrations, OAuth.
- Twitter v2
- WhatsApp Cloud
- IG Graph
- FB Pages
- TikTok
- Stripe
- Slack
-
Layer 02
Departments
Fifteen specialist teams. Each owns a domain and ships artifacts (drafts, reports, decisions).
- Customer Care
- Brand
- Legal
- Finance
- +11
-
Layer 03
Agent Fleet
Sixty-seven named specialist agents with personas, calibration histories, and tier promotions. Self-improving via the drill harness.
- Atlas
- Aiko
- Lena
- Yara
- Roman
- Janus
- Argus
- Sage
- Quinn
- Juana
- +57
-
Layer 04
Brain Core
The runtime — multi-tenant Postgres with row-level security, per-tenant DEK encryption, write-once audit log, drill harness.
- Supabase 17
- RLS-by-default
- Per-tenant DEK
- Write-once audit
- Drill harness
Brain core capabilities
What the runtime guarantees, not what it suggests.
Most AI products promise outcomes. BBB's Brain promises invariants — properties that hold whether you have one tenant or ten thousand.
-
Per-tenant DEK encryption
Every tenant's secrets sealed with a tenant-specific data encryption key. Wrapped by your master KEK, which you control.
-
Row-level security
Postgres RLS scoped to your org_id, JWT-driven. Same database, every customer, with provable isolation auditable in the schema.
-
Write-once audit log
Every state change writes a row. Updates and deletes are rejected at the SQL level. Export at any time for any regulator.
-
Idempotency by construction
Inbound keyed by (org, platform, external_id). Outbound by deterministic hash. Replay any event without producing duplicates.
-
Drill harness — self-improving
Daily drills, dual-graded, scored 0–10. Tier promotions at three consecutive 9+. Tier demotions at two consecutive ≤4.
-
Bilingual at the model layer
Drafter routes to Khaleeji-trained Arabic models for Arabic inbound; English models otherwise. No Google Translate in the loop.
Spin up your tenant in under a minute.
Database, agents, OAuth — all provisioned automatically. You're drafting your first reply within five minutes of signup.