Project work, managed hosting, and AI usage are priced as different things because they are different things.
Factory Sketch
Map one expensive workflow, identify what can safely move to agents, define the foreman role, and produce the first factory blueprint.
Workflow map
Candidate agent roles
Risk list
Review gates
Foreman loop
Build/no-build recommendation
Factory Build
Use the Singularity Labs factory to build a client-specific factory with roles, memory rules, dashboards, and handoff procedures.
Intake and specification flow
Agent role definitions
Execution plans
Verification gates
Observability trail
Operating runbook
Factory Foreman Training
Teach the human operator to steer the factory, diagnose confusion, approve risk, improve the loop, and stop the line when quality drops.
Foreman operating model
Specification habits
Review routines
Failure diagnosis
Memory hygiene
Escalation rules
Managed Agent Hosting
Host and maintain managed agent workbenches and client sites after launch: infrastructure, updates, usage logs, backup routines, credentials hygiene, and scoped access.
Managed runtime
Private client workbench
Agent access controls
Usage and activity logs
Maintenance routine
Billing handoff
Managed Factory Operation
Run a client-specific factory on a dedicated VPS or private runtime, then surface client-safe factory feedback through the portal.
Dedicated VPS or private runtime
Factory health dashboard
Feedback event timeline
Upkeep and backup routine
Foreman training status
Compute posture and caps
Factory Audit
Inspect brittle AI workflows that only work while one person babysits them, then add the machinery that makes them usable.
Brittleness review
Risk inventory
Missing checks
Recovery paths
Operator handoff gaps
Repair plan
The billing model should match the work.
A factory build is not the same thing as keeping a server alive, and neither one is the same thing as production model usage.
Build work is fixed-scope
Factory Sketch and Factory Build work are quoted as scoped projects. The quote is based on the shape of the problem, the risk, and the value of a working factory.
Hosting is recurring
If Singularity Labs runs infrastructure for you, site hosting, dedicated VPS costs, monitoring, updates, backups, credential hygiene, and upkeep are billed monthly.
AI usage is measured separately
Build-time experimentation is part of R&D. Production usage is estimated after the loop is stable enough to benchmark, then handled as prepaid, capped, or client-supplied compute.
Hourly is for transition work
Hourly work is still available for small support windows, but it is not the default for factory design. The goal is a concrete system, not a long block of time.
Managed factory operation
For clients who want Singularity Labs to keep the machinery running, we can provision and manage the runtime, connect billing, maintain the agent workspace, and give the client a controlled dashboard for requests, approvals, and usage evidence. That is a monthly operating service, not a one-time website launch.
Compute is not labor.
AI model usage is treated like a metered operating cost. If a client wants Singularity Labs to run production decisions, the portal tracks the agreed budget separately from the project fee. Early R&D is quoted as research work; production usage is quoted after the loop is repeatable enough to measure.
Start by finding the factory shape.
The first useful question is not the price. It is whether the work can be made inspectable, recoverable, and teachable enough for a human foreman to operate.