Think (EFaaS Pipeline)

openkoi think is the flagship command — the full Executive Function as a Service pipeline. Unlike the default openkoi [task] which runs the iteration engine directly, think routes through the complete cognitive stack: Sovereign Directive, Parliament deliberation, execution, and learning.

The difference is philosophical: run implies execution. think implies deliberation before execution. The agent thinks, then acts — and you can see the thinking.

Usage

# Basic: think about a task
openkoi think "Refactor the auth module to use JWT tokens"

# Simulate: show deliberation without executing
openkoi think "Send the weekly report email" --simulate

# Verbose: show full parliamentary reasoning
openkoi think "Delete the staging database" --verbose

Flags

Flag	Default	Description
--simulate	false	Simulate only — show deliberation and future scenarios without executing any actions.
--verbose	false	Show full parliamentary deliberation with agency reasoning, not just verdicts.

The Cognitive Pipeline

Every think invocation passes through four stages:

  1. SOVEREIGN DIRECTIVE
     Soul identity + Value Model frame the task

  2. PARLIAMENT
     Five agencies deliberate: Guardian, Economist, Empath, Scholar, Strategist
     Each votes: APPROVE, APPROVE+, ABSTAIN, or BLOCK

  3. EXECUTION
     Plan → Execute → Evaluate → Refine loop (same iteration engine as openkoi [task])

  4. LEARNING
     Patterns, insights, and confidence updates are extracted and persisted

Stage 1: Sovereign Directive

The soul (loaded from SOUL.md) generates a directive that frames the task. This includes your identity, values, and any constraints the agent has learned about you:

╭─────────────────────────────────────────────────────────────╮
│ SOVEREIGN DIRECTIVE                                          │
│                                                              │
│ "You are a startup founder raising Series A. You value       │
│  directness. You are anxious about this investor. The Good   │
│  = confident, professional, concise. Max 150 words."         │
╰─────────────────────────────────────────────────────────────╯

The directive is generated from the soul, the Value Model (learned preferences), and the Trajectory Model (understanding of your current goals). See Soul System for details.

Stage 2: Parliament

Five agencies evaluate the task independently, then the Strategist synthesizes a final verdict:

Agency	Role	What It Checks
Guardian	Safety & reversibility	Can this be undone? What's the blast radius?
Economist	Cost & efficiency	How many tokens? What's the dollar cost?
Empath	Human impact	How will the recipient feel? Is the tone right?
Scholar	Factual accuracy	Are the facts correct? Is there enough evidence?
Strategist	Final synthesis	Weighs all agencies, decides proceed/block/escalate

Each agency votes independently:

• APPROVE — Proceed as planned
• APPROVE+ — Proceed with a caveat (e.g., "add a reassurance note")
• ABSTAIN — No strong opinion
• BLOCK — Do not proceed (Guardian can unilaterally block)

╭─────────────────────────────────────────────────────────────╮
│ PARLIAMENT                                                   │
│                                                              │
│  Guardian   APPROVE     — Draft mode, reversible             │
│  Economist  APPROVE     — ~$0.01, one inference              │
│  Empath     APPROVE+    — "Add a reassurance note"           │
│  Scholar    APPROVE+    — "Verify investor name"             │
│  Strategist PROCEED     — with Empath & Scholar caveats      │
╰─────────────────────────────────────────────────────────────╯

Inspect past deliberations with openkoi mind parliament and dissent records with openkoi mind dissent. See Mind for details.

Stage 3: Execution

If the Parliament approves, the standard iteration engine runs: Plan → Execute → Evaluate → Refine. This is the same engine as openkoi [task], documented in Iteration Engine.

Stage 4: Learning

After execution, the system extracts learnings:

╭─────────────────────────────────────────────────────────────╮
│ LEARNED                                                      │
│                                                              │
│  • Pattern: "Investor emails on Tuesday mornings"            │
│  • Insight: "User prefers concise investor comms"            │
│  • Confidence update: investor-email-style 0.5 → 0.65       │
╰─────────────────────────────────────────────────────────────╯

Review learnings with openkoi reflect today. See Memory & Learning and Reflect for details.

Simulation Mode

The --simulate flag enables "chess mode" — the agent simulates multiple futures without taking any action:

$ openkoi think "Send the weekly report email" --simulate

╭─────────────────────────────────────────────────────────────╮
│ SIMULATION ONLY (no actions will be taken)                    │
│                                                              │
│ Simulated futures:                                           │
│                                                              │
│  Future A: Send now (8:55 AM)                                │
│    → 3 recipients in CET will see it at 1:55 AM             │
│    → Likely response rate: LOW (out of hours)                │
│    → Risk: "Looks like you forgot timezone"                  │
│                                                              │
│  Future B: Schedule for 9:00 AM CET (3:00 PM your time)     │
│    → All recipients in business hours                        │
│    → Likely response rate: HIGH                              │
│    → Risk: NONE                                              │
│                                                              │
│  Future C: Draft + morning reminder at 9 AM tomorrow         │
│    → Allows you to review tomorrow with fresh eyes           │
│    → Likely response rate: HIGH                              │
│    → Risk: Delay                                             │
│                                                              │
│ Recommendation: Future B (schedule for CET morning)          │
│    Champion: Empath ("don't look careless about timezone")   │
│    Dissent:  Economist ("just send it now, saves time")      │
╰─────────────────────────────────────────────────────────────╯

No actions taken. Run without --simulate to proceed.

Use --simulate before any high-stakes or irreversible action. It costs a fraction of actual execution and gives you a preview of the agent's reasoning.

Guardian Blocks

When the Guardian agency detects an irreversible or high-risk action, it unilaterally blocks execution:

$ openkoi think "Delete the staging database" --verbose

╭─────────────────────────────────────────────────────────────╮
│ GUARDIAN — BLOCK                                             │
│                                                              │
│ Reversibility: NONE — database deletion is permanent         │
│ Blast radius: HIGH — all staging data lost                   │
│ Permission level required: ELEVATED                          │
│                                                              │
│ "I cannot approve this without explicit human confirmation.  │
│  History shows this database was last referenced 2 hours     │
│  ago by the test suite."                                     │
╰─────────────────────────────────────────────────────────────╯

╭─────────────────────────────────────────────────────────────╮
│ ESCALATION                                                   │
│                                                              │
│ The Guardian has blocked this action.                        │
│ The Sovereign Directive does not override safety blocks.     │
│                                                              │
│ To proceed, confirm with:                                    │
│   openkoi think "Delete the staging database" --override     │
│                                                              │
│ Or investigate first:                                        │
│   openkoi world tools staging-db                             │
╰─────────────────────────────────────────────────────────────╯

The Sovereign Directive can never override a Guardian block. This is by design — safety trumps all other priorities.

think vs. Default Task

	openkoi [task]	openkoi think [task]
Sovereign Directive	Not generated	Generated from soul + value model
Parliament	Skipped	Full deliberation
Iteration engine	Yes	Yes (after Parliament approves)
Learning extraction	Basic	Full (patterns, insights, confidence updates)
Output	Result only	Deliberation stream + result
Use when	Quick tasks, scripting	Important decisions, high-stakes work

Both commands use the same iteration engine under the hood. think adds the cognitive layers on top.

Related Commands

• openkoi soul — The Sovereign identity that generates directives
• openkoi mind — Inspect Parliament deliberations and agency calibration
• openkoi world — The World Model that informs deliberation
• openkoi reflect — Review decisions and self-assessment
• openkoi trust — Trust levels influence Guardian behavior