Duo-node Logo
|
RJS-Software Labs Logo

How Omnipotent Orchestrates AI Agents

Omnipotent isn't a chatbot. It's a closed-loop runtime that perceives, plans, executes, and validates work.

Zero-Install: The Managed Box

We handle the complexity of the runtime so you don't have to. With our Managed Box, the entire orchestration engine runs in a dedicated cloud container.

📦

1. Unbox

Spin up a dedicated runtime instance instantly. No Docker or CLI setup required.

🔑

2. Authenticate

Securely provide your API keys (BYOK) via our vaulted dashboard.

🚀

3. Command

Input your high-level goal. The system manages the planning & output cycles automatically.

1. The Two-Loop Architecture

Human cognitive workflow separates Strategy (Planning) from Tactics (Execution). Omnipotent mirrors this with two distinct, interlocking loops.

1
Perceive

Read the environment: codebase, state file, blockers

2
Plan

Draft a journey with trials, assign agents, estimate mana

3
Execute

Dispatch trials to agents via the signal bus

4
Reflect

Evaluate outcomes, update memory, adjust strategy

1/4

The Planning Loop runs at the strategic level. Shadow Route perceives the project state, drafts a plan, dispatches it, then reflects on what worked.

2. The Planning Funnel

Ambiguous goals (“Build an app”) are dangerous for AI. Omnipotent forces a rigorous breakdown process before a single line of code is written.

Strategic Funnel

From “Prophecy” to “Trial”. Every strategic goal is broken down into executable atoms. The quality of your planning directly impacts the quality of your work: no exponential backoff, just "yin and yang" efficiency.

Trial dependency graph

Work enters the system from a Prophecy refined through the funnel, or from an existing epic adapted to the planning structure. Both converge at the Quest level.

ProphecyExisting EpicRefineAdaptQuestTrial ATrial BTrial CCheckpointQUALITY GATE
Prophecy (top-down) Existing Epic (lateral) Converge → Quest Checkpoint Gate

Click a node to trace its dependency chain

3. The 9-Step Protocol

Every task execution follows a strict, non-negotiable state machine. This prevents hallucination loops and ensures human oversight at critical checkpoints.

STEP 01

Assess

Read swarm state, task board, and active blockers.

STEP 02

Plan

Identify files, dependencies, and verify test commands.

STEP 03

Execute

Implement code changes. One task at a time.

STEP 04

Verify

Run domain-specific tests (Rust, Deno, Node).

STEP 05

Update

Update status docs, task board, and swarm memory.

STEP 06

Checkpoint

MANDATORY. Wait for human approval to proceed.

STEP 07

Dispatch

Assign unblocked tasks to other idle agents.

STEP 08

Recycle

Return to Step 1. Pick up next priority.

STEP 09

Integrate

Run full suite when phase is complete.