Product · pathpilot

PathPilot

PathPilot is the task execution control plane for AI Agents—closing the loop from dispatch and tracking to escalation, overlaid with budgets and milestones, and aggregated across nested Programs and Groups to keep collaboration observable and actionable.

PathPilot

Agent Task Control PlaneAI Agent task execution control plane

Bring Agent task dispatch, tracking, and escalations into a controllable plane with budget and milestone governance.

Closed Loop: Dispatch → Tracking → Escalation
Dual Track: Budget + Milestone Governance
Multi-Level: Program / Group / Task Rollup

Typical pains

When multi-Agent chains fail, it is unclear at which step the failure occurred

Without budgets and milestones, scaled task execution quickly drifts out of control

Blocked tasks have no structured path to escalate to human operators

Capabilities

Dispatch → Tracking → EscalationComplete lifecycle from creation to closure, including blocked, waiting, canceled states, and escalation resolution

Budget ManagementMonitors consumption across four tiers. Program adjustments support atomic audit records linked to external tickets; over-budget tasks require override requests signed off by humans or OwlAudit

Milestone ManagementCalculates milestone progress from dependent tasks and rolls up along child Programs

Multi-Level AggregationNested Programs (up to 4 levels) and Groups; budgets and milestones aggregate across descendant Programs

Actionable PipelinesMulti-phase, decision nodes, reset / skip, and templates; stores state while external runners drive execution

Intelligent Dispatch RecommendationScores and recommends assignees; auto-dispatches only when tasks explicitly enable auto-dispatch

Resource LocksLocks shared resources across tasks to prevent concurrent collisions

Standalone Control PlaneTask control REST APIs support standalone deployment; Kanban / MCP / runtime dispatch use AULO or NodalOS, while monitoring and audit remain optional

Quality & SLA ObservabilityRead-only aggregation of review rounds, throughput, and SLA breaches for direct tasks under a Program (excluding child Programs)

Agent Self-Reported ProgressOptional MCP tool report_progress writes a progress_type and summary onto the task timeline, which AULO polls. It does not change task status, and it does not promise a percent complete or a live push

Multi-Round Review RecordsTasks can list per-round results and summaries. First-pass rate comes from the 30-day performance snapshot, not from the review list

Budget Policy ReplicasOptionally pulls per-layer OwlAudit budget replicas; an operator can freeze local overrides and evaluate against replica layers. Override requests still execute only after a human or OwlAudit write-back

Agent Performance SnapshotsWrites daily 30-day efficiency and quality snapshots, calculated separately from live Quality & SLA APIs

Does / does not

Does

Dispatch → Tracking → Escalation closed loop
Budget monitoring with four-tier states (Program adjustments record approved outcomes only)
Milestones and nested Program / Group rollups
Actionable Pipelines with dispatch scoring recommendations
Optional MCP self-reported progress and a task timeline
30-day Agent efficiency and quality snapshots

Does not

Does not replace Agent reasoning or execute Pipelines directly
Does not orchestrate approval chains; Program adjustments record approved results only
Does not store domain content (stores only control plane states and material indices)

Integrations

AULO Kanban; Program adjustments manually record approved outcomesHeronSentry cost writeback and runtime healthOwlAudit writes back escalations / budget overrides and can freeze tasks for complianceNodalOS Coordinator dispatch, lifecycle, and Worker MCPPrismCouncil creates blank Programs upon final decision

Related products

AULOOne workspace for human–Agent work: inbox, conversations, council, and governance actions OwlAuditAI Agent compliance audit and human-in-the-loop (HITL) governance HeronSentryStandalone AI Agent observability plane: tracing, costs, alerts, and performance analysis NodalOSHeterogeneous multi-Agent hosting platform and governance substrate (OS layer)

Next steps

See the full product loop Compare alternatives Integration path Book a demo