Products

Buy the section that hurts. Add the rest when it pays.

Meltrix, Formeon, Anneon, Seedscan, Panebot and Glastwin form one closed perceive, decide, act loop across the glass plant. Each is a product on its own and an agent in the platform.

Book a plant review See pricing
Products

Six products, one runtime

Every product shares the plant-edge runtime, the model router, the connectors and the audit log. Each is sold on the Line tier where it fits a single line, and bundled on Plant and Enterprise.

Meltrix

agent.batch_melt · Batch-and-Melt · Line or Plant tier

Autonomous batching and furnace melting, held on chemistry every second.

Problem

Melting is where glass is won or lost. Batching raw materials and melting them at 1,500°C into a homogeneous, bubble-free glass sets the seed and stone count, the homogeneity and the energy bill for everything downstream. Most plants still run fixed furnace setpoints, spot lab checks and the craft of retiring glass technologists. A batch or melt drift silently seeds bubbles, stones and cords, burns extra fuel, and is only caught after forming, as cullet, rejects or a field escape.

What Meltrix does

Meltrix is the autonomous batch-and-melt agent. It senses batch weighing, furnace temperature, pull rate and melt homogeneity, predicts their downstream effect on seeds, stones, cords and grade using Seedscan defect signals and lab data, and continuously adapts batching, furnace temperature and pull to hold target glass chemistry at the lowest energy and cullet, under human-approved graduated autonomy.

Built for

Flat, container, display and solar-substrate, specialty and fiber-glass makers running continuous furnaces; batch-plant and furnace-control OEMs.

Key features

  • Real-time batch, melt-chemistry and homogeneity modelling with closed-loop furnace control
  • Energy-per-tonne and cullet optimisation to grade at the lowest fuel and carbon cost
  • Defect coupling: uses Seedscan seed, stone and cord risk to pre-empt melt upsets
  • Composition- and furnace-specific recipe learning with ware-genealogy logging
  • Validated against Glastwin before any furnace setpoint change goes live

NVIDIA stack

H100 / HGX training on multi-plant furnace telemetryJetson Orin plant-edge inferenceCUDA and cuDNN surrogate solvingTensorRT edge control modelsTriton serving per furnaceNeMo fine-tuned glass-technology reasoningNVIDIA AI Enterprise and NIM

MVP scope

Advisory melt-chemistry and energy optimisation on one furnace using existing lab and SCADA data; graduate to closed-loop furnace control under technologist approval.

Open: Furnace actuator and batch-plant interfaces vary by site. Physics-informed furnace-CFD twin coupling is planned, not built.

Formeon

agent.form_shape · Form-and-Shape · Line or Plant tier

Autonomous float-bath and IS-machine forming, dialled to exact dimension.

Problem

Forming turns a molten stream into a pane or a bottle at speed. Float-bath draw for flat glass and gob-and-IS-machine forming for containers decide thickness, flatness, dimension and shape. A gob-weight error, a pull-rate drift or a forming instability produces off-thickness, warped or misshapen ware: cullet, rejects and breakage found only at inspection. Today it runs on fixed forming setpoints and the craft of a retiring forming-operator workforce.

What Formeon does

Formeon is the autonomous form-and-shape agent. It senses gob weight, pull rate, float-bath conditions and formed dimension, predicts thickness, flatness and shape outcomes using Seedscan and dimensional signals, and continuously adapts forming setpoints and gob weight to hit exact dimension at rate, under human-approved graduated autonomy, on a homogeneous melt delivered by Meltrix.

Built for

Flat and automotive-glazing float lines; container-glass IS-machine plants; display, solar-substrate and specialty drawing or pressing lines; forming-machine OEMs.

Key features

  • Real-time gob-weight, pull-rate and dimension modelling with closed-loop forming control
  • Thickness, flatness and shape optimisation to grade at maximum takt
  • Defect coupling: uses Seedscan check and inclusion risk to pre-empt forming faults
  • Product- and machine-specific forming-recipe learning with ware genealogy
  • Validated against Glastwin before forming setpoint changes go live

NVIDIA stack

H100 / HGX training on multi-line forming telemetryJetson Orin line-edge inference for gob, pull and forming controlCUDA and cuDNN glass-flow surrogatesTensorRT deterministic edge latencyTriton forming and dimension endpointsNeMo process-reasoning modelsNVIDIA AI Enterprise and NIM

MVP scope

Advisory dimensional and thickness optimisation on one line using existing forming and dimensional data; graduate to closed-loop gob and pull control under operator approval.

Open: IS-machine and float actuator interfaces vary by site. Physics-informed glass-flow twin coupling is planned, not built.

Anneon

agent.anneal_stress · Anneal-and-Stress · Line or Plant tier

Autonomous lehr annealing that relieves stress before it becomes breakage.

Problem

Annealing decides whether glass survives. The lehr must cool formed glass on a precise curve to relieve residual stress. Get it wrong and the glass carries hidden stress and anisotropy that turn into breakage on the line, in tempering, or worst of all as a safety-critical field escape when a pane shatters or a bottle bursts. Most lehrs run fixed cooling profiles and spot polariscope checks, blind to real-time residual stress.

What Anneon does

Anneon is the autonomous anneal-and-stress agent. It senses lehr-zone temperatures, cooling rate and residual-stress and birefringence signals, predicts stress and breakage risk, and continuously adapts the lehr cooling curve to hit target residual stress with minimum energy, under human-approved graduated autonomy, on ware formed by Formeon.

Built for

Flat, architectural, automotive, container and specialty glass makers running annealing lehrs and downstream tempering; lehr and tempering-equipment OEMs.

Key features

  • Real-time lehr-zone and residual-stress modelling with closed-loop cooling control
  • Stress, anisotropy and breakage-risk prediction from optical and birefringence fusion
  • Energy-optimal annealing curves that hold stress spec at the lowest fuel cost
  • Product-specific annealing-recipe learning with ware genealogy and stress records
  • Validated against Glastwin annealing-stress simulation before curve changes go live

NVIDIA stack

H100 / HGX training on multi-plant lehr and stress telemetryJetson Orin line-edge inference at the lehrCUDA and cuDNN annealing-stress surrogatesTensorRT edge stress modelsHoloscan and DeepStream optical and thermal fusionTriton stress and cooling-control endpointsNVIDIA AI Enterprise and NIM

MVP scope

Advisory stress-risk and annealing-curve optimisation on one lehr using existing thermal and optical data; graduate to closed-loop cooling control under operator approval.

Open: Optical and birefringence sensor coverage varies by site. Physics-informed annealing-stress twin coupling is planned, not built.

Seedscan

agent.defect_inspect · Defect-and-Inspect · Line or Plant tier

See every seed, stone, cord and stress before it escapes the plant.

Problem

The faults that sink glass are small and fast: a seed, a stone, a cord, an inclusion, a check, an edge chip, a thickness or flatness drift, or hidden residual stress. On a container line running thousands of pieces an hour, or a float ribbon moving continuously, spot inspection and human eyes miss faults until they become cullet, customer rejects or a field escape.

What Seedscan does

Seedscan is the plant-edge defect-and-stress perception layer. It fuses camera, optical, thermal and stress or birefringence streams to detect and predict seeds, stones, cords, inclusions, checks, edge damage, thickness and flatness drift and residual stress at line speed, then feeds that signal to Meltrix, Formeon and Anneon so the plant acts before the next fault, not after.

Built for

Flat, architectural, automotive, container, display, solar-substrate and specialty glass makers; inspection-equipment OEMs embedding Seedscan-ready sensing.

Key features

  • Multi-sensor fusion of vision, optical, thermal and stress or birefringence at the edge
  • Real-time detection and prediction of seeds, stones, cords, inclusions, checks and stress
  • Thickness, flatness and dimension drift sensing tied to grade spec
  • Defect genealogy per ware, feeding the plant's assurance-grade quality audit log
  • Rare-defect coverage boosted with synthetic seed, stone, cord and stress scenarios

NVIDIA stack

Jetson Orin on 4 to 24 camera, optical and stress stations per lineH100 / HGX multimodal defect trainingMetropolis and DeepStream multi-camera pipelinesHoloscan low-latency sensor fusionTensorRT for 30 to 60 FPS and sub-100 ms alertsCosmos and Omniverse Replicator synthetic rare defectsTriton and NIM serving

MVP scope

Passive multi-sensor defect detection on one line with genealogy logging; graduate to predictive signals that drive Meltrix, Formeon and Anneon control.

Open: Camera, optical and stress station coverage and lighting vary by site. Jetson Thor and AI Enterprise packaging are planned, not built.

Panebot

agent.robot_handling · Robot-and-Handling · Plant tier

Autonomous handling of hot, fragile glass: panes, ware and stacks.

Problem

Glass is heavy, hot, sharp and fragile, and moving it is where people get hurt and ware gets broken. Panes come off the float line, containers off the IS machine and lehr, all needing pickup, transfer and stacking without a chip, a scratch or a break. Manual and fixed-automation handling cannot adapt to product changes, breaks fragile ware, and leans on a shrinking, injury-exposed labour pool that caps rate.

What Panebot does

Panebot is the autonomous robot-and-handling agent. Using vision-guided robots, it picks, transfers, orients and stacks panes, containers and substrates around hot and fragile glass, adapting grip and path to product, temperature and Seedscan defect signals so good ware is handled gently and flagged ware is diverted, under human-defined safety limits.

Built for

Flat and automotive glass makers stacking panes; container plants palletising ware; display and solar-substrate handlers; glass-handling and robotics OEMs.

Key features

  • Vision-guided robotic pick, transfer, orient and stack for panes, ware and substrates
  • Adaptive grip and path planning for hot, fragile and varied glass geometries
  • Defect coupling: diverts Seedscan-flagged ware and protects good ware
  • Robot-path validation in Glastwin and Isaac Sim before deployment on the line
  • Handling-event genealogy tied to the plant's quality audit log

NVIDIA stack

Jetson Orin on-robot perception and controlH100 / HGX grasp and motion-policy trainingIsaac ROS and Isaac ManipulatorIsaac Sim path validation around hot, fragile glassTensorRT on-robot latencyDeepStream multi-camera pick and placeNVIDIA AI Enterprise and Fleet Command

MVP scope

Vision-guided pick-and-stack on one product on one cell with simulation-validated paths; graduate to defect-aware routing and multi-product handling.

Open: Robot make, model and end-effector interfaces vary by site. Sim-to-real handling policies are planned pending pilot validation.

Glastwin

agent.twin · Glass-line twin · Plant tier

Hit target quality, stress and takt before the run, in simulation first.

Problem

Glass makers commit to a run and find out later if it worked. A grade change, a new composition, a forming or annealing tweak or a takt push can seed defects, stress or lost rate that only show up in cullet and rejects hours later. Furnace campaigns are long and product qualifications are unforgiving, so every unproven change on the real plant is expensive. Plants have no way to test the full melt-form-anneal chain before it runs.

What Glastwin does

Glastwin is the as-formed glass digital twin. It simulates batching, furnace melting, forming, annealing and inspection geometry to predict as-formed quality, residual stress and takt before the run, and auto-optimises batch, melt, forming and annealing setpoints with cuOpt so Meltrix, Formeon and Anneon act on validated, approved moves instead of guesses.

Built for

Flat, architectural, automotive, container, display, solar-substrate and specialty glass makers running grade changes and new compositions; furnace, forming and lehr OEMs.

Key features

  • GPU-accelerated furnace-CFD, glass-flow and annealing-stress simulation of the full chain
  • As-formed quality, residual-stress and takt prediction before the run
  • cuOpt auto-optimisation of batch, melt, forming and annealing setpoints within envelopes
  • Grade-change and new-composition sequencing to cut cullet and qualification time
  • Twin-validated autonomy: approved control moves are replayed before promotion

NVIDIA stack

DGX / OVX hosting the twin and GPU furnace simulationH100 / HGX surrogate training and large CFD runsOmniverse glass-line digital twinCUDA and cuDNN furnace-CFD, glass-flow and stress solvingcuOpt pull-rate, energy-window and sequence optimisationCosmos and Omniverse Replicator rare-scenario synthesisOmniverse Cloud and NVIDIA AI Enterprise

MVP scope

Offline twin for one line predicting as-formed quality and stress on grade changes; graduate to cuOpt auto-optimisation and twin-validated closed-loop promotion.

Open: Fidelity of site-specific furnace, forming and annealing models varies. Grace Hopper hosting and physics-informed residual blending are planned, not built.

Run timeline · scenario

Six products in one transition

Every Glasent run is an ordered, inspectable sequence. This is scenario_fl2_01 on FL-2 · float line · clear soda-lime: Thickness change FL-2 · 6 mm to 4 mm clear float, residual stress inside spec, zero escaped seeds. It is a worked scenario that shows the shape of a run, not a measured customer result.

scenario_fl2_01 · agent graph · FL-2 10 nodes · 1 approval gate · scenario
01 · orchestrator ingest.order ✓ SUCCEEDED 02 · twin twin.simulate ✓ SUCCEEDED 03 · batch_melt melt.pull ✓ SUCCEEDED 04 · form_shape form.ribbon ✓ SUCCEEDED 05 · anneal_stress anneal.curve ✓ SUCCEEDED 06 · defect_inspect inspect.ribbon ✓ SUCCEEDED 07 · orchestrator yield.balance ✓ SUCCEEDED 08 · orchestrator approve.human ◆ APPROVAL 09 · robot_handling handle.stack ✓ SUCCEEDED 10 · orchestrator ware.qualify ✓ SUCCEEDED
scenario_fl2_01 FL-2 · float line · clear soda-lime scenario SUCCEEDED
  1. 01 ingest.order SUCCEEDED 0.9 s

    Plant Orchestrator · Pulled the 4 mm clear float spec, optical-grade tolerances and the standing energy window from plant MES; locked the target envelope for the run.

  2. 02 twin.simulate SUCCEEDED 41 s

    Glastwin · Simulated 36 candidate transition recipes across furnace pull, tin-bath ribbon speed and lehr curve; ranked them on seed risk, residual stress and energy per tonne.

  3. 03 melt.pull SUCCEEDED 6 m 20 s

    Meltrix · Stepped furnace pull toward the new ribbon mass flow while holding melt temperature and fining; chemistry stayed inside the composition window.

  4. 04 form.ribbon SUCCEEDED 4 m 05 s

    Formeon · Raised ribbon speed and re-angled the top rollers to thin the ribbon toward 4 mm; thickness converged inside the design tolerance.

  5. 05 anneal.curve SUCCEEDED 3 m 12 s

    Anneon · Re-shaped the lehr cooling curve for the thinner, faster ribbon so residual stress stays inside spec at the higher speed.

  6. 06 inspect.ribbon SUCCEEDED continuous

    Seedscan · 8 camera, optical and stress stations streaming; a seed cluster flagged at the ribbon edge and attributed to the pull transient, routed to cullet.

  7. 07 yield.balance SUCCEEDED 2.4 s

    Plant Orchestrator · Re-sequenced cut sizes so transition ribbon routes to cullet recovery and good ribbon to the highest-value open order.

  8. 08 approve.human APPROVAL 48 s

    Plant Orchestrator · The second pull step exceeded the site autonomy threshold. Held for the glass technologist on shift; approved and written to the audit log.

  9. 09 handle.stack SUCCEEDED 1 m 10 s

    Panebot · Re-planned pick and stack paths for the thinner panes; plates flagged by Seedscan diverted to cullet, good plates stacked to rack A3.

  10. 10 ware.qualify SUCCEEDED 1 m 02 s

    Plant Orchestrator · Lot released with full genealogy: batch, melt, forming, lehr curve, defect map, stress map and the technologist's approval.

Compare

Which product is in which tier

Line covers one module on one forming or annealing line. Plant covers every product and the additional scopes across the site. Enterprise adds custom models and a fleet.

Products by tier
ProductLinePlantEnterprise
MeltrixOne lineIncludedIncluded + custom models
FormeonOne lineIncludedIncluded + custom models
AnneonOne lineIncludedIncluded + custom models
SeedscanOne lineIncludedIncluded + custom models
PanebotNot includedIncludedIncluded + custom models
GlastwinNot includedIncludedIncluded + custom models
Coat-and-Temper, Quality, Yield and Takt, Glass-KnowledgeNot includedIncludedIncluded
Indicative price$16,000 / line / mo$95,000 / mo$700k – $9M ACV
Tool calls · scenario

One log, every product

Every tool invocation, argument and result is written to an immutable, human-readable log, and every reasoning step is expandable. Nothing about a run is hidden from the plant.

tool-call stream · scenario_fl2_01
  1. 02:14:03mes.read_orders({ line: "FL-2", next: "CLR-4MM" })
  2. 02:14:03ok spec locked · 4.0 mm · optical grade · stress limit set
  3. 02:14:04twin.simulate({ candidates: 36, horizon: "22 min" })
  4. 02:14:45ok best recipe #19 · seed risk low · stress inside spec
  5. 02:14:46policy.evaluate({ recipe: 19, autonomy: "L3" })
  6. 02:14:46ok 11 writes permitted · 1 write requires human approval
  7. 02:14:47scada.write_setpoint({ tag: "F1.PULL_SP", step: 1 })
  8. 02:21:07ok pull step 1 settled · fining stable · chemistry in window
  9. 02:21:08float.write_ribbon({ speed: "+", rollers: "re-angle" })
  10. 02:25:13ok thickness converging to 4.0 mm · flatness in tolerance
  11. 02:25:14lehr.write_curve({ zones: 12, profile: "4mm-fast" })
  12. 02:28:26ok residual stress inside spec at new belt speed
  13. 02:28:27vision.stream({ stations: 8, model: "seedscan-v3" })
  14. 02:28:41seed cluster · ribbon edge · attributed to pull transient
  15. 02:28:42cuopt.sequence({ objective: "cullet+energy" })
  16. 02:28:44ok transition ribbon routed to cullet · good ribbon to order
  17. 02:28:45approval.request({ action: "pull_step_2" })
  18. 02:29:33ok approved by the technologist on shift · logged
  19. 02:31:02isaac.plan_path({ stack: "A3", thickness: "4.0 mm" })
  20. 02:32:12ok paths validated in Isaac Sim · flagged plates diverted
  21. 02:33:15mes.log_lot({ lot: "FL2-4MM", genealogy: "full" })
  22. 02:34:17ok scenario run complete · stress in spec · 0 escaped seeds

Reasoning trace

Plan, thought, action, observation, exactly as the orchestrator would record it.

GoalMove FL-2 from 6 mm to 4 mm clear float with residual stress i…

Move FL-2 from 6 mm to 4 mm clear float with residual stress inside spec and zero escaped seeds.

PlanSimulate the transition on the twin first. Lead with furnace p…

Simulate the transition on the twin first. Lead with furnace pull so the ribbon mass flow is right before the forming change, then thin the ribbon, then re-shape the lehr curve. Watch the ribbon continuously; hold the second pull step for a human.

ThoughtThe standing recipe changes pull and ribbon speed together. On…

The standing recipe changes pull and ribbon speed together. On the twin that produces a fining transient that seeds the ribbon edge. Sequencing the pull step ahead of the forming change and holding fining removes most of it.

Actiontwin.simulate(candidates=36) returned recipe #19: pull in two …

twin.simulate(candidates=36) returned recipe #19: pull in two steps, forming change after fining settles, lehr curve re-shaped before belt speed rises.

ObservationRecipe #19 scored lowest on seed risk with stress inside spec.…

Recipe #19 scored lowest on seed risk with stress inside spec. Two candidates scored lower on energy but pushed residual stress over the site limit and were discarded.

ActionExecute recipe #19 under autonomy level L3: eleven setpoint wr…

Execute recipe #19 under autonomy level L3: eleven setpoint writes permitted, the second pull step routed to the glass technologist.

ObservationSeed cluster at the ribbon edge at 02:28:41, attributed to the…

Seed cluster at the ribbon edge at 02:28:41, attributed to the pull transient. cuOpt routed that ribbon to cullet recovery; no flagged plate reached a customer stack.

OutcomeScenario run complete. Thickness at 4.0 mm, residual stress in…

Scenario run complete. Thickness at 4.0 mm, residual stress inside spec, one approval gate, full genealogy written to the lot record.

How it works

A goal becomes a plan becomes a run

Nothing about a Glasent run is implicit. The goal is declared, the plan is simulated on the twin, each step is typed, and the outcome is written to the ware record.

Active path

The orchestrator plans, the agents act

A run starts as a goal, becomes a plan, and executes as a graph of typed steps. The active path is always visible, always logged, always reversible.

RUNNING APPROVAL SUCCEEDED
< 100 ms Defect alert on high-speed container lines · design target
30–60 FPS Vision throughput per station · design target
4–24 Camera, optical and stress stations per line
< 500 ms Fused stress and flatness decision on flat glass · design target
Human in the loop

Autonomy is a dial, not a switch

Shadow, assist, bounded, then unattended inside an envelope. Each site sets the level per agent, per tag, per shift, and every write above the threshold waits for a named glass technologist.

Module paths

Typical first modules

Where plants usually start, and what they add when the first number is proved.

Inspection first

Seedscan, then Meltrix

Start passive: detect and attribute seeds, stones and checks with genealogy. Add melt control once attribution shows the furnace is the cause.

Forming first

Formeon, then Anneon

Start on gob weight or ribbon thickness in advisory mode. Add the lehr curve when stress at the cold end is the next number.

Annealing first

Anneon, then Seedscan

Start on residual stress and breakage risk on one lehr. Add inspection fusion to attribute checks to their cause.

Simulation first

Glastwin, then the control agents

Start offline: predict as-formed quality and stress on grade changes. Promote validated recipes to Meltrix, Formeon and Anneon as the plant gains confidence.

Autonomy

Four levels, set per agent and per tag

A plant does not go from manual to unattended in one step. Glasent makes the level explicit, auditable and reversible at any time, and the first release plan is shadow, then assist, then graduated autonomy.

Autonomy levels and the human role at each
LevelWhat the agent doesWhat the person doesWhen
L1 · Shadow and advisoryObserves, predicts and recommends setpoints with its reasoningEnters every change manually; a baseline is measuredPilot weeks 1 to 3
L2 · AssistProposes a write; it executes on approvalApproves each write in the review consolePilot weeks 4 to 8
L3 · BoundedWrites inside tag, rate and magnitude limits on low-risk loopsApproves pull steps, grade releases and anything above thresholdPilot week 9 onward
L4 · UnattendedRuns the approved envelope without promptingSets the envelope; reviews the shift recordPlanned, after graduated autonomy proves out
Accelerated computing

Physical AI, at the furnace

A float ribbon never stops and an IS machine forms thousands of pieces an hour. Perception has to be local, deterministic and fast, so Glasent runs GPU inference at the plant edge and keeps training, simulation and optimisation on DGX, HGX and OVX.

Jetson Orin · DeepStream · Holoscan · TensorRT

Edge perception

4 to 24 synchronised camera, optical, thermal and stress stations per line. Sub-100 ms defect alerts on container lines and sub-500 ms fused stress and flatness decisions on flat glass are the design targets.

Triton · NIM

Model serving

Defect, stress, time-series drift and process-risk models served across edge and plant servers, with glass-knowledge and reasoning endpoints packaged as NIM services.

DGX / HGX · NeMo

Training

Multimodal models fine-tuned on inspection imagery, optical and stress outputs, PLC time series, recipes, gob weight, furnace temperature, pull rate, lehr curves, energy, cullet and final grade. Planned cadence: monthly plant refreshes.

Omniverse · OVX

Digital twin

GPU-accelerated furnace CFD, glass-flow and annealing-stress simulation of the as-run line, with 10 to 100 candidate recipes evaluated per grade change.

Omniverse Replicator · Cosmos

Synthetic data

50,000 to 250,000 rare seed, stone, cord, inclusion, check and stress scenes per glass family, always validated against real inspection distributions before training use.

cuOpt · RAPIDS

Optimisation

Pull-rate constraints, energy windows, cullet routing, forming and annealing sequence and line takt, with RAPIDS for high-volume telemetry ETL.

Measured

Design targets, stated as targets

These are the numbers the architecture is built to hit and the pilot is built to measure. None is a customer result yet; every pilot report reproduces its figures from the plant's own ware genealogy.

90–120 days Paid line pilot, shadow to assist to bounded writes
3–5 Design-partner plants in the first cohort
50k–250k Synthetic rare-defect scenes per glass family · planned
1,500°C Where the process starts, in furnaces that run for a decade
Developers

Define an agent, bound it, run it

The Glasent SDK is typed Python. Tools are declared with schemas and limits; the policy engine enforces them at call time, not in a review meeting.

What you get

  • Typed tool definitions with unit-aware ranges and rate limits
  • Deterministic replay of any historical run against a new model
  • Local twin harness so a recipe is simulated before it is shipped
  • Autonomy policy as code, versioned and reviewed like any other change

Read the docs Developer guide

fl2_anneal_agent.py 
# Bound the anneal agent to twelve lehr zones on FL-2.
from glasent import Agent, Tool, Limit, Autonomy

lehr = Tool(
    name="lehr.write_curve",
    tags=["FL2.LEHR.Z01..Z12.TEMP_SP"],
    limits=[Limit(max_step="4 C", per="60s")],
)

anneal = Agent(
    id="agent.anneal_stress",
    goal="residual stress inside spec, min energy",
    tools=[lehr, Tool("optic.read_birefringence", read_only=True)],
    # bounded writes; a technologist still gates pull steps
    autonomy=Autonomy.L3,
    # simulate on Glastwin before every write
    verify="twin",
)

run = anneal.start(line="FL-2", product="CLR-4MM")
for step in run.stream():
    print(step.name, step.status, step.duration)
From the plant floor

The problems we hear

No customer quotes yet; we are pre-launch. These are the three buyer personas the product is built for and the pain each one describes, in their own terms.

"The furnace has run for years on the same setpoints. The process never has. We find out a melt drifted when the cullet pile grows."

Plant / operations director · ICP persona

"I can tell you why a check appeared from the lehr curve and the gob weight. I cannot be at every line, and the people who could are retiring."

Glass technologist · ICP persona

"A missed seed is a reject. A missed stress fault is a pane that shatters in the field. I need genealogy on every piece, not a spot check."

Quality / reliability engineer · ICP persona
Integrations

It speaks plant, not cloud

Glasent reads and writes through the furnace, forming, lehr, inspection and MES systems already on the floor. No rip-and-replace, no parallel historian, no new HMI to learn.

Furnace and batch

Furnace SCADA and PLC, batch-plant weighing, redox and fining instruments

Setpoint reads and guarded writes over OPC UA and Modbus

Forming

Float-bath and IS-machine controls, gob-weight and timing systems

Gob, pull, ribbon speed and roller reads; guarded writes

Annealing

Lehr zone controllers and belt drives

Zone temperature and curve reads; guarded writes

Inspection

Camera, optical, thermal, polariscope and birefringence stations

Frames, stress maps, defect records, line-speed streams

Glass MES

Orders, grades, lots and ware genealogy

Spec and tolerance reads; genealogy and release writes

Historian

Time-series stores and lab information systems

Backfill, replay and lab chemistry

Robotics

Robot cells, conveyors and stackers via NVIDIA Isaac

Pick, path and stack commands inside the safety envelope

Identity and edge

SSO and RBAC via SAML or OIDC; NVIDIA Jetson Orin edge nodes

Named approvers, tag-level roles, sub-100 ms inference

See all integrations

Guardrails

An agent that can move a furnace needs a leash

Glasent writes to production equipment. Every capability is scoped, every write is policy-checked, and every action is written to an append-only audit log the plant owns.

  • Bounded action space. Each agent can only write to an explicit tag allow-list, inside per-tag rate and magnitude limits.
  • Policy engine before every write. Autonomy level, shift, product, interlock state and operator presence are all evaluated before a setpoint moves.
  • Human-in-the-loop gates. Anything above the site threshold, pull steps, grade releases, safety-adjacent moves, waits for a named approver.
  • Immutable audit log. Append-only, hash-chained, exportable, and retained on the plant's own storage.
  • Hard fallback. Loss of the edge node, the network or the model returns control to the furnace, forming and lehr systems' last known-good state.
  • Tenant and IP isolation. Compositions, forming recipes and defect libraries never cross a customer boundary. On-prem deployment available.

Compliance posture

StandardScopeStatus
SOC 2 Type ICloud control plane RUNNING Planned in the first six months
SOC 2 Type IICloud control plane QUEUED Planned in months six to twelve
IEC 62443Plant-edge OT security RUNNING Design-aligned
ISO 9001 / IATF 16949Quality and genealogy records SUCCEEDED Record formats supported
Container and safety-glass standardsStress and defect conformance records SUCCEEDED Record formats supported

Read the security overview

Enterprise

Standardise autonomy across the group

One policy model, one audit trail, one benchmark across every line in every plant, with the composition and forming models kept private to each site.

Talk to us Enterprise details

  • SSO and role-based access down to the tag level
  • Group-wide autonomy policy with per-site override and approval chains
  • Cross-plant benchmarking on cullet, seeds, energy per tonne and first-pass quality
  • VPC, on-prem and air-gapped plant-edge deployment options
  • Custom composition, forming and tolerance models per site
  • Dedicated deployment engineer and quarterly model review per plant
FAQ

Straight answers

The questions plant directors and glass technologists ask in the first meeting.

Yes, but only within an explicit tag allow-list with per-tag rate and magnitude limits, and only at the autonomy level your site has set. Every pilot starts in shadow mode, where Glasent predicts and recommends and a person enters everything. Writes come later, after the recommendations have earned it.

Get started

Pick your first product

One product on one line with a pre-agreed baseline. The pilot report tells you whether the second one is worth it.

Book a plant review See pricing