OMEGA (Orchestrated Multi-Expert Governed Agents) is a governed execution platform for coordinating generative, deterministic, and policy-driven agents with built-in proof, privacy, and failure semantics.
- OMEGA Governed Execution Whitepaper (Markdown): whitepapers/omega-governed-execution.md
- OMEGA Governed Execution Whitepaper (PDF): assets/whitepapers/omega-governed-execution.pdf
A production-grade execution substrate for orchestrating multiple classes of agents across tools, workflows, and environments. Learn more
To make delegation, automation, and agentic execution safe, explainable, and defensible by default. Read the core concepts
OMEGA treats execution as something that must be provable, reconstructable, and constrained by design. See the full analysis
Routing, execution, memory, workflow coordination, extensibility, and agent collaboration as first-class primitives. Explore capabilities
OMEGA’s core enforces the complete lifecycle of autonomous digital entities — from creation through termination — with explicit authority, receipt-backed finality, and policy-driven automation.
- Governed Birth (PT-014 — Proven)
- Entity creation requires explicit human or system authority
- Receipt-bound birth events with fail-closed semantics
- No entity exists without a governed genesis record
- Governed Death (PT-016 — Proven)
- Revocation and termination produce immutable lineage
- No death without birth (prevents phantom entities)
- No double-death (prevents state corruption)
- Two death modes: policy revocation vs system termination
- Receipt chains link birth → death with no gaps
- Governed Automation (PT-017 — Proven)
- Policies can trigger automatic revocation or termination
- Severity gradation: RECOMMEND → AUTO_REVOKE → AUTO_TERMINATE
- Human gate enforcement for irreversible actions
- Cooldown periods prevent policy flapping
- Full attribution: policy ID, version, automation flag, trigger events
- Fail-closed: ambiguous cases default to NO_ACTION
- Constitutional Guarantee
- Machines may act automatically — but always prove why, under whose authority, and with what limits
- Every automated action is attributable, auditable, and reversible up to the point of execution
- Receipt chains survive forensic reconstruction months or years later
Every execution can emit a deterministic receipt that proves what happened without exposing content. Understand receipts
Privacy is enforced structurally through schema, hashing, and policy boundaries rather than convention. Best practices
Same inputs produce the same proofs, enabling replay, verification, and auditability. Proof semantics
Keon provides cryptographic governance and audit guarantees for OMEGA’s execution layer without changing its execution model. Architecture overview
Once execution is provable, downstream systems become inevitable rather than aspirational. See deliverables
A modular system composed of agents, execution engines, routing layers, memory systems, and optional governance substrates. Full architecture
OMEGA exposes execution through a deliberately constrained user interface called the Workshop.
The Workshop allows users to:
- discover workflows
- initiate governed execution
- observe run state and governance interruptions
It does not make decisions or render evidence.
- Workshop UI (Execution Surface)
- Governed Execution Diagram
- Auditor Walkthrough
- Why Not Open Source (Yet)
- Separation of Powers
How does governed execution differ from conventional agent platforms? This framework evaluates systems on architecture, not claims.
OMEGA can be adopted incrementally, from local execution to governed multi-tenant workflows. Quick start guide
Detailed documentation covering the complete capability set, advanced systems, and deliverables. Complete documentation
Restraint, clarity, credibility, and correctness over novelty and unchecked autonomy. Advanced systems reference
Every claim in OMEGA is backed by a sealed, tagged, independently verifiable proof campaign. These are not demos — they are immutable evidence bundles with SHA256-hashed manifests.
| Campaign | What It Proves | Status |
|---|---|---|
| PT-003 | Agent Registry & Capability Routing | ✅ Proven |
| PT-004 | Workflow Orchestration Execution Spine | ✅ Proven |
| PT-005 | Explicit GATE_REQUIRED Ergonomics (Human-in-the-Loop) | ✅ Proven |
| PT-013 | Multi-Titan Collaboration under FC Governance | ✅ Proven |
| PT-014 | Genesis under Human-Governed Execution (Birth) | ✅ Proven |
| PT-016 | Governed Revocation & Death Semantics (Death) | ✅ Proven |
| PT-017 | Revocation Policy Automation (Automated Governance) | ✅ Proven |
Birth, Death, and Automation are all governed, attributed, and provable.
Full proof artifacts, harness code, and evidence bundles: OMEGA Proof Campaign Status
An actively developed, production-tested system with 7 sealed proof campaigns covering the full governance lifecycle. Documentation status
Focused on deepening proof semantics, federation flexibility, and system evolvability. Roadmap details
Teams building systems where correctness, explainability, and control matter. Start here
Toy agents, prompt demos, or systems optimized solely for novelty.
Built to be extended, audited, and evolved through disciplined collaboration.
Every execution produces a deterministic receipt. Here's a concrete example:
{
"event": "email.action.executed",
"workflowId": "wf_personal_signal_firewall",
"stepId": "act.reply_decline",
"status": "success",
"correlationId": "corr_...",
"inputHash": "sha256:...",
"outputHash": "sha256:...",
"receiptHash": "sha256:...",
"meta": { "source": "gmail", "fromDomain": "recruiterco.com", "reasonCodes": ["stack_mismatch"] }
}