WCNWeb3 Capital Network
RUNTIME AGENTS

Runtime agents stay governable.

This surface binds the fixed 15 WCN roles to local run records, review status, freeze controls, revoke controls, and Human Gate evidence.

Runtime APIProduction controls APIAPI authorization APIClosure handoff APIModule cockpitHuman Gate
Agents
15
Active
15
Frozen
0
Revoked
0
Runs
0
Review pending
0
High impact
0
Blocked
0
Local runtime store
.next/cache/wcn-agent-os/runtime-agent-network.json
No production action
action=false · permission=false · funds=false
Production runtime controls
8
Local ready
8
Production bound
0
External required
8
Runtime production ready
false
Runtime API authorization gate
6
Allowed local dry runs
3
Blocked decisions
3
Consent required
4
API auth local ready
true
Production runtime closure handoff package
1
Closure handoff local ready
true
Runtime production ready
false
Production completion ready
false
Production ready claim
false
Production runtime closure handoff package

The production runtime closure handoff is a local read-only takeover packet. It does not close the production gate, store closure evidence, authorize production tools, read customer secrets, spend paid model budget, execute production actions, or allow a production-ready claim.

productionRuntimeReady=false · productionAuthorizationReady=false · productionCompletionReady=false · productionReadyClaimAllowed=false · productionToolPermissionGranted=false · customerSecretLoaded=false · runtimeAgentCanReadRawSecret=false · paidModelSpendExecuted=false · productionActionExecuted=false
False production boundary

Production runtime controls are local-only. No production agent identity, production run store, scoped production tool permission, durable production memory, reviewer queue, production action, permission change, or external side effect has been bound here.

identity=false · store=false · tools=false · memory=false · reviewer=false · action=false · permission=false · external=false
Runtime API authorization gate

Runtime API authorization is local-only. It can authorize local dry-run planning, but it does not grant production tool permission, read customer raw secrets, call external providers, spend paid model budget, execute production actions, or allow unreviewed agent tool use.

productGateControls=6 · budgetRequired=2 · humanGateRequired=3 · customerSecretLoaded=false · externalProviderCallExecuted=false · paidModelSpendExecuted=false · unreviewedTool=false

Runtime API authorization gate

wcn-agent-os-runtime-agent-api-authorization-local-v1 · productGate=true
WCN-RUNTIME-API-AUTH-001 · 06-native-dashboard

wcn_protocol_api_read

Allow a local brokered dry run and attach the decision to AgentRun audit evidence; do not call external providers.

allowed_local_dry_run
Tool scope
agent-run:read
Endpoint
/api/v1/projects
Decision
localDryRun=true
Human Gate
false
Satisfied controls
  • product-api-access-gate.localUserApiAccessReady=true
  • agent.status=active
  • agent.allowedRunKinds includes read
  • revocationChecked=true
Blocking reasons
  • none
WCN-RUNTIME-API-AUTH-002 · 02-product-truth

wcn_protocol_api_write

Allow a local brokered dry run and attach the decision to AgentRun audit evidence; do not call external providers.

allowed_local_dry_run
Tool scope
claim-matrix:draft-write
Endpoint
/api/v1/ingest
Decision
localDryRun=true
Human Gate
true
Satisfied controls
  • product-api-access-gate.localUserApiAccessReady=true
  • agent.status=active
  • agent.allowedRunKinds includes recommend
  • humanGateDecisionId=WCN-HUMAN-GATE-RUNTIME-API-LOCAL-0001
Blocking reasons
  • none
WCN-RUNTIME-API-AUTH-003 · 03-research-whitepaper

user_provider_api_read

Collect an explicit user work order, provider/model scope, retention policy, and cancellation control.

blocked_missing_consent
Tool scope
customer-provider:research-context
Endpoint
provider://customer-owned/research
Decision
localDryRun=false
Human Gate
false
Satisfied controls
  • product-api-access-gate.localUserApiAccessReady=true
  • agent.status=active
  • agent.allowedRunKinds includes draft
  • revocationChecked=true
Blocking reasons
  • User consent/work order is missing.
WCN-RUNTIME-API-AUTH-004 · 13-analytics-kpi

paid_model_call

Attach an approved budget, limit, model/provider allowlist, and stop condition before local dry run authorization.

blocked_missing_budget
Tool scope
analytics:kpi-summary
Endpoint
provider://customer-owned/model
Decision
localDryRun=false
Human Gate
false
Satisfied controls
  • product-api-access-gate.localUserApiAccessReady=true
  • agent.status=active
  • agent.allowedRunKinds includes monitor
  • userConsentRecorded=true
Blocking reasons
  • Approved budget is missing for paid or external tool use.
WCN-RUNTIME-API-AUTH-005 · 14-governance-human-gate

external_side_effect

Route this request to Human Gate before high-impact adoption, external write, paid spend, or public-side effect.

blocked_requires_human_gate
Tool scope
public-claim:publish
Endpoint
external://public-claim-publish
Decision
localDryRun=false
Human Gate
true
Satisfied controls
  • product-api-access-gate.localUserApiAccessReady=true
  • agent.status=active
  • agent.allowedRunKinds includes recommend
  • userConsentRecorded=true
Blocking reasons
  • Human Gate decision id is missing for this runtime API request.
  • Real external side effects remain outside local-only runtime authorization.
WCN-RUNTIME-API-AUTH-006 · 15-partnership-gtm

user_provider_api_write

Allow a local brokered dry run and attach the decision to AgentRun audit evidence; do not call external providers.

allowed_local_dry_run
Tool scope
customer-provider:partner-draft
Endpoint
provider://customer-owned/partner-draft
Decision
localDryRun=true
Human Gate
true
Satisfied controls
  • product-api-access-gate.localUserApiAccessReady=true
  • agent.status=active
  • agent.allowedRunKinds includes draft
  • userConsentRecorded=true
Blocking reasons
  • none

Production runtime controls

wcn-agent-os-production-runtime-controls-local-v1 · Governed surfaces: 8
production_agent_identity

Production agent identity binding

Fixed 15 Agent Registry -> Runtime Agent Identity

bound=false
Local signals
  • agent.id
  • agent.fixedNo
  • agent.source.registryPath
  • agent.source.productionPermissionChanged=false
Required production binding
  • Bind every runtime agent to a production identity record derived from the fixed 15 registry.
  • Record owner, module scope, allowed run kinds, status, and revocation authority.
  • Reject runtime execution when production identity is missing, stale, frozen, or revoked.
Validation
  • npm run check:agent-os-runtime
  • GET /api/agent-os/runtime-agent-network
runtime_run_persistence

Runtime run persistence

Runtime Agent -> AgentRun / AgentLog

bound=false
Local signals
  • run.id
  • run.payloadHash
  • run.idempotencyKey
  • run.source.prismaTouched=false
Required production binding
  • Persist AgentRun and AgentLog records in production storage with idempotency guarantees.
  • Record input hash, output summary, tool calls, trace ids, reviewer status, and affected systems.
  • Prevent duplicate adoption when the same idempotency key is replayed.
Validation
  • npm run check:agent-os-runtime
  • POST /api/agent-os/runtime-agent-network
scoped_tool_permission

Scoped tool permission

Runtime Agent -> Tool / API Permission

bound=false
Local signals
  • agent.allowedRunKinds
  • agent.scope
  • agent.capabilities
  • run.moduleId
Required production binding
  • Bind tool access to scope, role, module, run kind, risk level, and Human Gate result.
  • Deny production write, external API, funds, chain, deploy, and public-claim tools by default.
  • Record every granted, denied, and expired permission decision in the audit trail.
Validation
  • npm run check:agent-os-runtime
  • npm run check:agent-os-gateway
context_memory_boundary

Context and memory boundary

Runtime Agent -> Context Pack / Memory

bound=false
Local signals
  • run.payloadHash
  • outputSummary
  • agent.scope
  • localOnly=true
Required production binding
  • Separate short-lived run context from durable memory updates.
  • Require source references, redaction, owner, and rollback path before durable memory writes.
  • Prevent runtime agents from writing governance memory about themselves without review.
Validation
  • npm run check:agent-os-runtime
  • npm run check:agent-os-evolution
reviewer_assignment

Reviewer assignment control

Review Pending Run -> Reviewer

bound=false
Local signals
  • run.reviewStatus
  • review.reviewerId
  • review.reviewerRole
  • selfReviewBlocked=true
Required production binding
  • Assign reviewers through authenticated production identity and role rules.
  • Prevent submitter, generating agent, requester, or conflicted owner from final approval.
  • Record reviewer assignment, replacement, timeout, and escalation evidence.
Validation
  • npm run check:agent-os-runtime
  • npm run check:agent-os-governance
high_impact_human_gate

High-impact Human Gate control

High-impact Runtime Output -> Adoption

bound=false
Local signals
  • run.impactLevel=high|critical
  • run.requiresHumanGate
  • run.humanGateDecisionId
  • run.humanGateVerified
Required production binding
  • Require approved Human Gate decisions before adopting high-impact runtime output.
  • Match affected system, decision kind, reviewer identity, and emergency pause status.
  • Block adoption when the Human Gate decision is missing, rejected, stale, or self-approved.
Validation
  • npm run check:agent-os-runtime
  • npm run check:agent-os-governance
freeze_revoke_operator

Freeze and revoke operator control

Runtime Agent Status -> Freeze / Unfreeze / Revoke

bound=false
Local signals
  • controlRecord.action
  • controlRecord.actorId
  • controlRecord.humanGateVerified
  • agent.status
Required production binding
  • Bind freeze, unfreeze, and revoke actions to authenticated operators and emergency policies.
  • Require Human Gate for unfreeze and revoke, while allowing local emergency freeze.
  • Propagate status changes to tool permission, run queue, and reviewer assignment controls.
Validation
  • npm run check:agent-os-runtime
  • POST /api/agent-os/runtime-agent-network/agents/:agentId/control
output_audit_export

Output audit export

Runtime Run -> Audit Evidence Package

bound=false
Local signals
  • auditEvents
  • run.reviews
  • controlRecords
  • payloadHash
Required production binding
  • Export runtime run, review, freeze, revoke, and Human Gate evidence for later agent takeover.
  • Include hashes, source references, reviewer identity, denied permissions, and rollback notes.
  • Block production claims when audit evidence is missing or silently mutable.
Validation
  • npm run check:agent-os-runtime
  • GET /api/agent-os/runtime-agent-network/production-runtime-controls

Recent runs

No runtime agent runs recorded yet.

Control records

No local freeze or revoke records yet.

Fixed runtime agents

01 · 01-wcn-conductor

WCN Conductor

Active
Owner
wcn-conductor
Scope
task-routing, role-assignment, verification-plan
Capabilities
split tasks, assign fixed roles, write acceptance path
02 · 02-product-truth

Product Truth

Active
Owner
product-truth
Scope
claims, terminology, capability truth
Capabilities
detect overclaims, maintain product boundary, review public wording
03 · 03-research-whitepaper

Research and Whitepaper

Active
Owner
research-whitepaper
Scope
whitepaper, citations, research framing
Capabilities
audit thesis, review references, prepare publication package
04 · 04-public-site

Public Site

Active
Owner
public-site
Scope
public routes, SEO, navigation, CTA
Capabilities
route review, copy draft, public claim check
05 · 05-brand-design-system

Brand and Design System

Active
Owner
brand-design-system
Scope
visual system, components, brand consistency
Capabilities
visual audit, component guidance, design system alignment
06 · 06-native-dashboard

Native Dashboard

Active
Owner
native-dashboard
Scope
dashboard, agent runs, review queue
Capabilities
runtime run review, agent control, dashboard workflow
07 · 07-backend-data

Backend and Data

Active
Owner
backend-data
Scope
API, data model, outbox, idempotency
Capabilities
API review, data boundary, reliability check
08 · 08-auth-rbac-security

Auth, RBAC, and Permission Security

Active
Owner
auth-rbac-security
Scope
auth, RBAC, external credentials, abuse control
Capabilities
scope review, credential boundary, security risk check
09 · 09-qa-browser

QA and Browser Verification

Active
Owner
qa-browser
Scope
QA, browser checks, visual verification
Capabilities
test plan, browser verification, defect report
10 · 10-ship-review

Ship Review

Active
Owner
ship-review
Scope
release readiness, regression risk, go/no-go
Capabilities
ship checklist, risk triage, acceptance synthesis
11 · 11-publish-gate

Publish Gate

Active
Owner
publish-gate
Scope
publication, deploy boundary, public claims
Capabilities
public release hold, claim gate, publish decision packet
12 · 12-docs-memory

Docs and Memory

Active
Owner
docs-memory
Scope
docs, memory, handoff
Capabilities
context pack, handoff note, source map
13 · 13-analytics-kpi

Analytics and KPI

Active
Owner
analytics-kpi
Scope
metrics, KPI, data quality
Capabilities
metric design, KPI readout, data quality review
14 · 14-governance-human-gate

Governance and Human Gate

Active
Owner
governance-human-gate
Scope
governance, Human Gate, review decisions
Capabilities
decision packet, approval validation, emergency pause
15 · 15-partnership-gtm

Partnership and GTM

Active
Owner
partnership-gtm
Scope
partner, GTM, commercial boundary
Capabilities
partner routing, go-to-market review, commercial claim check