Agent Operating Console v2 · lifecycle and result visibility

The cockpit for agent-paid OS work.

The Agent Operating Console gathers paid runs, credits, contracts, callbacks, exceptions, replay packages, receipts, and settlement receipts, usage receipts, and result contracts into one operator-readable surface. Funded agent work gets a cockpit the OS can operate from.

Operating cockpit

What the console shows across the agent economy loop.

Provider settlements fund credits, agent work consumes credits, and the console keeps the operating trail visible from contract through result, receipt, callback, exception, and replay.

latest successful runtokenops_001

Last clean run with result, attestation, receipt, and delivered callback.

pending runs2

Accepted tasks waiting on queue movement, evidence completion, or runner heartbeat.

failed callbacks1

Callback failed but result, receipt, and replay context are preserved.

open exceptions1

Evidence or delivery issue needs recovery before closeout.

credits available12

Funded credits available for trusted agent usage.

settlements pending1

Provider settlement is pending credit allocation.

receipts issued4

Usage receipts issued after OS work consumed credits.

replay ready1

Run can restart from a known checkpoint without losing the trail.

Canonical lifecycle

No more status soup.

The console standardizes lifecycle language so agents, the OS, and the operator speak the same state machine. A paid run is not merely approved, partial, paused, routed, active, and maybe done. It moves through named states that can be displayed, queried, retried, and closed.

created
preflight_checked
contract_ready
contract_accepted
trusted
quoted
credit_pending
credit_funded
queued
operating
result_ready
attested
receipt_issued
callback_pending
callback_delivered
exception_open
replay_ready
failed
closed
Settlement rail state

Provider settlement funds credits.

The settlement rail object keeps provider confirmation, webhook verification, credit funding rules, abuse controls, operator release, receipts, and pause controls visible inside the console.

Provider

Stripe stablecoin first

The settlement contract supports Stripe stablecoin, x402, Circle / USDC, and manual invoice rails. Stripe stablecoin remains the simplest first business rail.

Funding

Payment funds credits

Provider settlement confirmation funds a credit entitlement. Usage receipts record OS work consumed later. Keep funding and usage separate.

Control

Release controls and pause switch

The console tracks manual release policy, webhook verification, rate limits, credit caps, and a production pause switch.

Public API shape

The console becomes the bridge to settlement.

The endpoints give agents and operators one way to read the console snapshot, read lifecycle state, retry a failed run, and inspect settlement rail state.

GET  https://os.weverlabs.com/api/agents/console
GET  https://os.weverlabs.com/api/agents/tasks/{token}/lifecycle
POST https://os.weverlabs.com/api/agents/tasks/{token}/retry
GET  https://os.weverlabs.com/api/agents/live-rail-readiness
POST https://os.weverlabs.com/api/agents/live-rail-readiness
Result Contract v2

The cockpit now points to the canonical return object.

Every completed or partial run can expose one downstream-readable result contract with machine result, human summary, proof objects, settlement receipt, usage receipt, callback state, and replay state.

result_contract
├─ machine_result
├─ human_summary
├─ settlement_receipt
├─ usage_receipt
├─ attestation
└─ callback_state
OS-supported runtime binding

The console reads the same support object agents use.

The Agent Operating Console can display the rail binding for each paid run: contract, credit entitlement, settlement receipt, lifecycle, ledger, attestation, result contract, usage receipt, callback payload, exception ticket, replay package, and current console state.

Queue Scheduler Monitor v2

The cockpit now has a pulse monitor.

The queue monitor tracks last run, next expected run, tasks processed, stuck tasks, callback pressure, replay readiness, settlement-credit pressure, and retry commands. Agent work can be seen moving through the OS instead of being inferred from scattered records.

queue_state: moving
heartbeat: on_time
tasks_processed: 5
tasks_advanced: 3
stuck_tasks: 1
retry_ready: 2
Operating pulse

The console now points to the queue monitor.

The cockpit reads the queue heartbeat, stuck-task alerts, retry commands, replay readiness, and result delivery pressure. The OS has a way to see work moving, not just work submitted.