LoopX Control Plane for Long-Running Work

Overview

LoopX is a provider-neutral control plane for long-running agent work. The series moves from durable Goal state and work graphs to governed turns, accountable checkpoints, monitors, backoff, replanning, self-repair, and domain extensions. Its central thesis is that continuity comes from owned identity, scoped authority, fresh evidence, validated writeback, and explicit next states—not from preserving a transcript or maximizing activity. LoopX can make lifecycle reliability inspectable across sessions, agents, hosts, and external systems, but it cannot guarantee model quality, domain correctness, or safe autonomous production.

Learning outcomes

LoopX preserves a Goal, work frontier, authority boundaries, evidence, and next legal transition beyond a session.
A runnable Todo requires more than open status: dependencies, scoped authority, capability, workspace, claims, and freshness also matter.
Validated writeback precedes quota accounting; waiting and unchanged monitor polls are not delivery progress.
Public evidence supports evaluating LoopX's control-plane design and bounded use cases, not general intelligence or production autonomy.

Keywords

LoopXagentsGoalTodoevidenceauthoritymonitorsbackoffprovidersrecovery

Chapters

LoopX: Why Sessions Need a Control Plane

Durable continuity without autonomy theater

11 min · Durable continuity without autonomy theater

11 min

LoopX State: Facts That Survive Sessions

Canonical state, projections, and local-first boundaries

12 min · Canonical state, projections, and local-first boundaries

12 min

LoopX Work Graph: Who May Act?

Dependencies, authority, capability, and workspace filters

11 min · Dependencies, authority, capability, and workspace filters

11 min

LoopX Governed Turn: When May It Act?

From quota decision to validated writeback

11 min · From quota decision to validated writeback

11 min

LoopX Checkpoint: From Goal to Action

Linking frontier, contract, evidence, and writeback

10 min · Linking frontier, contract, evidence, and writeback

10 min

LoopX Monitors: Wait Without Spinning

Backoff, scheduler hints, and host readback

10 min · Backoff, scheduler hints, and host readback

10 min

LoopX Stalled Work: Replan or Repair

Evidence deltas, retries, and honest closure

11 min · Evidence deltas, retries, and honest closure

11 min

LoopX Capabilities: From Kernel to Products

Providers, extensions, hosts, and adoption boundaries

15 min · Providers, extensions, hosts, and adoption boundaries

15 min

LoopX Reliability: Claims and Limits

Evidence, safety, liveness, and closure

12 min · Evidence, safety, liveness, and closure

12 min

Preview every chapter · Unlock full text, audio, and video from chapter 4

Plans

Ch.1-3 FreeCh.4-9 Premium

FAQ

What does the LoopX control plane do for long-running agents?

LoopX preserves a project-owned Goal, filters runnable work, governs bounded turns, and records evidence and next state across sessions.

When is LoopX worth using instead of a normal agent session?

LoopX is worth the added state and contracts when work crosses sessions, waits for external results, changes owners, or needs scoped authority and recovery. Small, closed tasks that are cheap to redo may need only a normal session.

How does LoopX handle waiting and stalled work?

LoopX represents external waiting with a target, due time, evidence, and termination condition, then uses backoff; stalled work is classified as waiting, retry, replan, self-repair, blocking, or closure.

What can LoopX prove about agent reliability?

LoopX can make lifecycle control, attribution, evidence, writeback, and recovery inspectable; it cannot by itself prove that a model chose the best route or that a domain result is correct.