Private by design · Human governed
Your company,
intelligently orchestrated.
Luminary AI brings your agents, approvals, integrations, and execution history into one self-hosted command center—without turning control over to a black box.
Runs on your infrastructure
◇ Approval-gated actions
Tamper-evident audit
luminary.local
COMMAND CENTER
Welcome, Administrator
A live view of your Luminary AI workspace.
!
Reply to customer emailProposed by Inbox Agent
MEDIUM!
Update CRM recordProposed by Sales Agent
LOW
✓
HUMAN APPROVALDecision recorded
◉
LOCAL INFERENCEYour data stays yours
Beta interface preview using demo data.
Private beta
Join the shortlist
Early access is for operators who want local AI leverage without giving up governance, deployment control, or auditability.
01 — THE IDEA
AI should expand your leverage.
Not erase your oversight.
Luminary is being built for teams that want useful automation and a clear answer to: what ran, why did it run, what data did it use, and who approved the result?
AI WORKFORCEConfigured agents
Reviews invoices and financial operations.
ACTIVEMEDIUM RISKSupports pipeline and customer follow-up.
ACTIVELOW RISKTriages messages and drafts safe proposals.
ACTIVELOW RISKHUMAN IN THE LOOPPending approvals
HIGH RISK · DEMOApprove external action
Policy requires a recorded human decision before this action can proceed.
MEDIUM RISK · DEMOReview email draft
Draft content is bound to the approval request and remains unsent.
AUDIT LEDGERVerified activityDEMO
✓ Approval recordedAdmin10:42VERIFIED
↻ Brief completedLocal Agent10:41VERIFIED
◇ Proposal createdInbox Agent10:40VERIFIED
⌁ Metadata importedSystem10:39VERIFIED
02 — FIRST BETA WORKFLOW
Email operations,
without the black box.
The first wedge is private email operations: bring Gmail and Outlook metadata into one governed inbox, support manual demo import before OAuth, run metadata-only triage locally, draft proposed responses, require approval, and preserve the trace.
01 Unified inbox ingest
02 Local/manual analysis
03 Approval-bound draft
04 Audited outcome
Useful before it becomes huge.
Instead of trying to automate the whole company on day one, Luminary starts with one workflow that almost every operator understands: the inbox. The staging build now supports unified inbox metadata, manual pre-OAuth testing, metadata-only triage with persisted results, and approval-bound draft proposals before anything touches an external system.
03 — THE GOVERNED WORKSPACE
One quiet place to run
the intelligent company.
Purpose-built surfaces turn AI activity into visible, governable work.
01
A workforce you can actually see.
Define specialist agents by department, capability, and risk. Every run becomes a persistent record—not a disappearing chat.
- Seeded and custom agent roles
- Durable runs plus read-only local/offload briefs
- Local Ollama inference foundation
AI WORKFORCEConfigured agents
Reviews invoices and financial operations.
ACTIVEMEDIUM RISKSupports pipeline and customer follow-up.
ACTIVELOW RISKTriages messages and drafts safe proposals.
ACTIVELOW RISK
02
Automation with a brake pedal.
Consequential actions stop at a human decision point. Approvals are bound to the exact proposed content and recorded before execution.
- Risk-tiered approval policy
- Separation of duties and quorum support
- Exactly-once execution leases
HUMAN IN THE LOOPPending approvals
HIGH RISK · JUST NOWSend vendor payment
Policy requires a recorded human decision before this action can proceed.
MEDIUM RISK · 12 MIN AGOReply to customer email
Draft content is cryptographically bound to this request.
03
Memory with receipts.
Trace activity across runs, approvals, integrations, and a hash-chained audit ledger designed to make silent changes visible.
- Append-only audit history
- Signed exports and integrity verification
- Tenant-scoped records and role checks
AUDIT LEDGERVerified activityCHAIN HEALTHY
✓ Approval grantedAdministrator10:42:18VERIFIED
↻ Run completedInbox Agent10:42:17VERIFIED
◇ Proposal createdInbox Agent10:41:52VERIFIED
⌁ Provider synchronizedSystem10:40:06VERIFIED
04 — BUILT DIFFERENTLY
Private infrastructure.
Deliberate autonomy.
Self-hosted
Run Luminary on hardware and infrastructure you control. The local development path binds to loopback by default.
Governed
Agents can propose. Policy and people decide. High-impact actions stay visible and interruptible.
Connected carefully
Provider integrations begin read-only, with mutation capabilities deliberately gated behind explicit controls.
Observable
Health, readiness, diagnostics, structured logs, backup, restore, and rollback are part of the product foundation.
05 — CURRENT BUILD
An honest staging product,
moving toward daily utility.
Luminary already has a serious security and operations foundation, with MFA, custom agents, unified inbox, local/offload worker, approval, and audit surfaces now wired in staging. The next milestone is closing the loop from real provider data to approved external action and a complete trace.
✓Working foundationAuthentication, tenants, approvals, audit, security controls, deployment tooling
↗Beta workflow in progressEmail triage loop, encrypted provider credentials, production integration wiring
○Not yetBroad autonomous company workflows or write-enabled provider actions
06 — WHO IT IS FOR
Built for operators
who need leverage and control.
The first buyers are likely privacy-conscious founders, agencies, internal ops teams, and technical service businesses with repetitive work and real approval risk.
Founder-led teams
Replace scattered tools and one-off automations with a governed company command layer.
Service businesses
Handle customer messages, proposals, follow-ups, and operational tasks without losing accountability.
Regulated workflows
Keep approvals, audit trails, and deployment boundaries visible from the first day.
07 — PRIVATE BETA
Not a waitlist.
A focused build circle.
Early access is intentionally small so the product can be shaped around real workflows without exposing source code or rushing unfinished automation into production.
Workflow fit call
We map the repetitive work, approval risk, data boundaries, and deployment constraints that matter to your team.
Private demo
You see the current staging build, the first email workflow direction, and exactly what is production-ready versus still under construction.
Pilot decision
If there is a strong fit, the next step is a narrow pilot around one real workflow with clear human approval and audit requirements.
THE COMPANY OS
Build leverage.
Keep the keys.
Luminary AI is under active development. Private beta conversations are open for teams that want local-first AI operations with human approval built in.
- Best for technical operators comfortable with private deployment
- Initial focus: email and operations workflows with human approval
- Commercial private beta — not open-source, freeware, or self-serve SaaS