A private agent workforce for your company.
Stood up inside your tools and run with CKS. Senior people in charge. It covers product, engineering, infrastructure, and intelligence, on the same platform that runs our whole company. We run on Neural. So can you.
What it is
Not a chatbot you ask questions. A workforce that does scheduled, accountable work in your stack, the way a good engineering team does. Agents handle the mechanical steps; people handle the judgment.
The workforce
Four functions, one orchestration.
Product & Delivery
Turns intent into organized, ready-to-build work: intake, scoping, specifications, tickets, and tracking.
Engineering
Takes ready work through a build pipeline: planning, implementation, and review, with senior engineers accountable for what ships.
Infrastructure & Ops
Keeps software alive: deployments, monitoring, security checks, backups, and on-call awareness.
Intelligence
Turns activity into decisions: research, analysis, status, and reporting.
A coordination layer routes work between the functions and keeps them consistent. People sit above the whole thing.
The registry
A roster, not a black box.
Every agent in the deployment has a name, a scope, and a human checkpoint. This is the shape of the fleet: what each one does, what it can reach, and where a person signs off.
doesRoutes work between agents, holds shared context, keeps the fleet consistent.
reachesThe coordination layer only
checkpointEscalates anything ambiguous to a person
doesIntake, triage, scoping, specs, tickets, board upkeep.
reachesTicketing & docs
checkpointA person approves scope before build
doesPlans, implements, tests, documents, prepares releases.
reachesRepos & CI, scoped per project
checkpointA senior engineer reviews every merge
doesDeploys, monitors, backs up, runs security sweeps.
reachesInfrastructure, least privilege
checkpointFindings reported; fixes approved by a person
doesResearch, analysis, status, recurring reporting.
reachesRead-mostly, across the stack
checkpointA person signs off on every conclusion that leaves the building
doesBriefings, inbox & calendar triage, drafts, reminders.
reachesOne person's accounts, private to them
checkpointNothing is sent without approval
The registry is the shape, not the ceiling. Deployments add agents per function as your operation needs them.
Proof of work
Five workflows we actually run.
Not hypotheticals. These run CKS today. Every workflow has the same shape: a trigger, agent work, a human checkpoint, an artifact.
The morning brief
Trigger
Every morning, before the day starts.
Agent actions
Reads calendar, inbox, project state, and open commitments; ranks what actually matters.
Human checkpoint
A person reads it. Corrections teach it.
Output
One prioritized brief.
Inbox triage
Trigger
Mail arrives.
Agent actions
Classifies, de-duplicates, and routes; drafts replies for the ones worth answering.
Human checkpoint
No email is sent without a person's approval.
Output
A triaged inbox and ready drafts.
The build pipeline
Trigger
A ticket is marked ready.
Agent actions
Plans the change, implements it, writes tests and docs; an independent pass reviews the work.
Human checkpoint
A senior engineer reviews and merges.
Output
A reviewed, merged change.
Ops watch
Trigger
On a recurring schedule.
Agent actions
Audits infrastructure, dependencies, and exposure; verifies backups actually restore.
Human checkpoint
Findings are reported, never silently fixed.
Output
A written audit with recommendations.
Board autopilot
Trigger
Daily, as the board drains.
Agent actions
Validates queued work against a quality gate, ranks it, promotes what's genuinely ready.
Human checkpoint
Humans set the gate. Anything ambiguous is flagged, not forced.
Output
A stocked, ranked board.
This is how the board stays stocked and priorities stay current: on a schedule, all of it visible.
People stay in charge. The system enforces it.
Governance here isn't a policy document. It's engineering. Seven rules, built into the platform:
Every agent runs on an explicit allowlist. If a capability isn't granted, it doesn't exist.
Each agent reaches only the systems its role requires. Least privilege is configuration, not policy.
Every material action is recorded with the agent that took it. You can always see what was done and by which agent. Attribution is built in, not reconstructed.
Anything consequential (an email, a merge, a deployment) stops and waits for a person, on boundaries you set. Routine background work runs on its own; nothing unsupervised makes consequential changes.
Snapshots precede anything destructive. Changes are reversible, and the way back is part of the design.
A single deployment inside your own tools and environment, not a shared service. Your code and data stay yours. Nothing of yours trains anyone else's model.
You can watch the activity, revoke access, and review or delete retained context whenever you want.
Every autonomous action is attributable, reviewable, and reversible.
Getting started
We stand it up and run it with you.
We wire it into your existing chat, ticketing, code, and deployment tools, load it with your context and conventions, set the approval boundaries with you, and operate it as a managed service, with senior CKS people accountable alongside yours.
It's built to fit how your team already works, not to force a new process on it.
Neural One
The same engine, sized down to a private chief of staff: briefings, inbox, goals, and a memory that answers to you.
See where agents would fit.
A 30-minute call. Tell us how your team ships and we'll show you where Neural Ops earns its place.