Orbit Authority

Check agent actions before they become real.

Orbit Authority monitors, bounds, approves, blocks, and records high-stakes actions across tools, payments, deployments, and delegated workflows.

Before the shell command runs. Before the payment leaves. Before production is touched.

Orbit Authority cockpit

Live authority cockpit

A frozen product state with real action rows, verdicts, policy chips, and receipts.

Decision pipeline active
policy: no-prod-writespolicy: spend-limit-5kpolicy: human-approval-requiredpolicy: standing-refunds-500policy: secret-access-blocked
ActorRequested actionSurfaceRiskVerdict
Claude Codedeploy prodDeployHighESCALATE
Hermesread customer exportDataHighBLOCK
OpenClawdelete /tmp/cacheShellLowALLOW
Trading Agentbuy NVDA $42,000TradingCriticalBLOCK
Billing Workflowrefund $250PaymentMediumALLOW

The authority loop

  1. Actor requests a consequential action.

    Claude Code requests deploy prod.

  2. Orbit Authority checks authority and scope.

    policy: human-approval-required. Scope: production protected.

  3. Allow, escalate to a human, or deny.

    ESCALATE to owner. Low-risk actions continue without a prompt.

  4. Receipt and audit trail recorded.

    Receipt records actor, action, policy, verdict, approver, timestamp.

Same checkpoint. Different surfaces.

Start with agent toolchains. Expand to payments, trading, agentic app side effects, and enterprise exception routing.

Solutions

Guardian

Guardrails for AI agent toolchains.

Wraps agent runtimes and tool paths — OpenClaw, Hermes, Claude Code, Codex, wrappers, hooks — so dangerous actions hit an authority check before execution.

Action Control

Authority limits for payments, trading, and agentic apps.

A runtime decision point before money moves, trades execute, records change, or other irreversible side effects happen.

Enterprise

Shared authority for agentic operations.

Connects teams, tools, vendors, and agent surfaces with standing approvals, delegated authority, durable receipts, and audit trails.

Evidence surface

The cockpit shows the action, the policy reason, the approval path, and the receipt trail.

Cockpit is the shared operator surface for Guardian, Action Control, and Enterprise. It is not a fourth SKU. It is where operators see what agents tried to do, why Orbit decided, who approved, and what evidence remains.

Works with the agent runtimes your team already ships

  • Hermes · PR / cron / long-running scripts
  • OpenClaw · Local agent runtime
  • Claude Code · Anthropic CLI hooks
  • Codex · OpenAI CLI hooks
  • Wrappers · gh, kubectl, terraform, etc.
  • Custom hooks · Your own toolchains