NEW MAP

CivOS

V0.9
DraftStatus
0Sources
0Risks
A flight simulator for real-world decisions

Test-fly a strategy before it meets the real world.

Bring CivOS an uncertain idea. It builds the flight plan—goal, people, constraints, evidence, and time horizons—then specialist Agents simulate pressure, failure modes, and possible outcomes before you commit.

About 2 minutes to set up · No recommendation is generated until you confirm the decision state.
01Test a strategy

Turn a rough proposal into a reversible experiment.

02See who is affected

Surface costs, benefits, missing voices, and conflicts.

03Explore across time

Compare immediate feasibility with longer-term effects.

Live agent run · observed workflowPreparing decision traceUnavailable tools are marked explicitly. Stage times reflect this live run.
0.0sReady
Manager orchestrationDecision Council
Waiting for specialist outputs
ManagerWaiting to synthesize

Specialist results must arrive before the Manager can conclude.

EventSelect a trace node
Input
—
Tool / method
—
Output
—
Stage time
—
Voice DeliberationTalk through the decision

Speak naturally. CivOS will extract the goal, constraints, and affected stakeholders—not hold a voice chat.

Add your evidence or notes
Multimodal civic input0 sources

Attach real working material. CivOS classifies each item and marks it as User-provided in the decision graph.

PDFs and images use multimodal analysis when OpenAI is connected. Binary files are not saved in Decision Memory.

Add a decision to create an AI-assisted draft.

Why this direction

A reversible pilot creates evidence before a larger commitment.

Key assumptions
  • The problem is urgent enough to justify a pilot.
  • An owner and success metric can be agreed.
Counterpoint

A pilot can still consume trust and attention if the decision criteria are unclear.

Guardrail ReceiptTrust Receipt
Pending analysis
  • No unsupported sources presented as verified
  • Personally sensitive data excluded
  • External actions require approval
  • Minority-impact check completed
  • Unresolved assumptions disclosed
Receipt not issued
Visible safety evidenceGuardrail Receipt
Command paletteWhat do you want to inspect?⌘K / Ctrl+K
Press Esc to close
Model connectionHow should CivOS run?

Managed AI is the easiest option. A personal OpenAI key is optional and stays only in this open page.

Managed connection selected.
Technical deep diveTrace / Eval DashboardNo trace
Span timelineObserved wall-clock durations
GuardrailsReceipt results
Eval rubric0/8 checks
Sensitive trace payloads are excluded; dashboard shows operational metadata only.
Human Approval Gate · External action boundary CivOS recommends consulting affected residents before finalizing this plan.

Approve preparation of an outreach draft?

No outreach will be prepared or sent until you decide.
Issue Your decision map
Continue exploringNot ready for a pilot or full rollout
0 of 5 checks complete
Evidence strengthLow
Unknowns3 unresolved
ReversibilityEasy
Decision readinessExplore
Disagreement0 material conflicts
Decision briefUntitled decision
Inside CivOS

From a question to testable possibilities

CivOS clarifies the decision, convenes independent specialist perspectives, simulates materially different outcomes across stakeholders and time, then returns one traceable decision state—while consequential actions remain under human control.

Observed agent architectureHow a decision becomes a reviewable model
Solid arrows pass structured outputs · external actions require approval
InputDecision + files
TriageClarifying Agent

Ask one material question

OrchestratorDecision Manager
SpecialistOperator
SpecialistSkeptic
SpecialistStakeholder
SpecialistEvidence Auditor
Scenario toolsWeb / File SearchRisk guardrails
Outcome simulationPossibility ranges

Stakeholder impacts · time · warning signals

Decision stateStructured & versioned

Assumptions · evidence · conflicts

Reviewable resultDecision Map + Brief
Only when an external or consequential action is requested
Authorization boundaryHuman Approval Gate
Scoped executionExport · Save · External Actions
01

Clarify before acting

One high-value question anchors the strategy, affected people, constraints, and time horizon before Agents run.

02

Independent council

Operator, Skeptic, Stakeholder, and Evidence Auditor submit separate structured findings to one Manager.

03

Simulate distributions

Multiple outcome ranges show who may benefit or bear burdens, what drives each case, and which signals change it.

04

Human approval

The system may investigate and prepare, but external or consequential actions stop for explicit approval.

Now · V0.9Adaptive intake, Live Agent Run, independent Decision Council
CurrentStakeholder-aware Outcome Simulation with explicit uncertainty ranges
CurrentEvidence provenance, decision memory, scoped approval gates, Action Tools