Request access
Private Preview

Fantasti Cloud Platform Private Preview

Every layer of a GPU cloud, on one account.

The Fantasti Cloud Platform is compute, development environments, ways to run work, storage and networking, and the tools to operate them. Every product shares one API, one set of policies, one quota system and one statement.

Request access Read the API
fx = fantasti.Client()

Opens with the first cohort. No account is open yet.

TAB 01One account Illustrative
Record 6 fields
account
acme
products
20 · one client Products: §01 Compute: 5 (4 Private Preview, 1 By request). Develop: 3 (2 Early access, 1 Planned). Data and network: 2 (2 Private Preview). Run: 3 (1 Private Preview, 2 Early access). Operate: 4 (1 Private Preview, 2 Early access, 1 Planned). Access: 3 (2 Private Preview, 1 Early access).
policy
research.yaml · applies to every request Shared controls: §06
capacity
on-demand · spot · reserved Capacity: §03
quota
reviewed by the Orchestrator Quota: Orchestrator
statement
one · split by cost center One bill: Pricing
Interfaces
Python · REST · CLI · MCP
NVIDIA GPUs
6 models · 2 rack-⁠scale systems
Metering
per second · one statement
Stage
Private Preview · requests open
§01 Products 20 products · 5 pillars

What is on the platform.

§02 How it works Reviewed 2026-10-10

Capacity underneath.
A cluster that is yours.

A capacity layer of NVIDIA GPU nodes underneath. On it, an isolated cluster for each account, running the tools you already use. You work through one API, one console and one bill.

How the Fantasti Cloud Platform works. Illustrative. Reviewed 2026-10-10.

Illustrative. Six products reach the platform through one API: GPU Clusters (Private Preview), Workspaces (Early access), Sandboxes (Early access), Flows (Planned), Batch Inference (Private Preview), Serverless (Early access). The figure follows one call, fx.clusters.create(gpu="H200:8", nodes=2). It arrives as the request POST /v1/clusters. The Fantasti Orchestrator checks it at five gates, each with the value the request carries: GPU H200:8 × 2, fabric infiniband, region us-*, capacity spot, max price ≤ 3.10. It is placed on the capacity layer, on us-east, fabric a, as cluster cl_8f2k: 2 nodes, 16 dedicated GPUs, one InfiniBand fabric and a control plane of your own. us-east, fabric b is passed over: 8 free < 16. The same layer holds rack-scale systems (72 GPUs in one NVLink domain, by request), and CPU Instances, Storage, Networking, Managed services. Usage is metered per second onto one statement for account acme, on the cost center research-llm. Connections: Workspaces to POST /v1/clusters; Sandboxes to POST /v1/clusters; Flows to POST /v1/clusters: One API; Batch Inference to POST /v1/clusters; Serverless to POST /v1/clusters; GPU Clusters to POST /v1/clusters; POST /v1/clusters to H200:8 × 2; H200:8 × 2 to infiniband; infiniband to us-*; us-* to spot; spot to ≤ 3.10; ≤ 3.10 to cl_8f2k: Placed; cl_8f2k to acme: Metered.

Fantasti OrchestratorBuilt to be run by agents
Capacity layerGPU capacity, storage and networks
Fabric aInfiniBand · us-east
  • GPU Clusters (highlighted)
    fx.clusters.create( gpu="H200:8", nodes=2)
    Private Preview
  • Workspaces
    fx.workspaces.create( gpu="H100:1")
    Early access
  • Sandboxes
    fx.sandboxes.create( ttl="15m")
    Early access
  • Flows (planned)
    @fantasti( gpu="H200:8", capacity="spot")
    Planned
  • Batch Inference
    fx.batch.create( input="s3://…")
    Private Preview
  • Serverless
    fx.jobs.create( image=…, gpu="H200:8")
    Early access
  • RequestPOST /v1/clustersREST
    { "name": "sft-01", "gpu": "H200:8", "nodes": 2, "capacity": { "mode": "spot", "max_price": "3.10" } }
  • GPUH200:8 × 2 (ok)
  • Fabricinfiniband (ok)
  • Regionus-* (ok)
  • Capacityspot (ok)
  • Max price≤ 3.10 (ok)
  • Your clustercl_8f2kIsolated
    HGX H200 8-GPU baseboard, line drawing8 GPUs per node
    nodes
    2 × H200:8
    fabric
    one · InfiniBand
    control
    your own plane
    state
    running
    16 GPUs, dedicated
  • On the same layer
    • CPU Instances
    • Storage
    • Networking
    • Managed services

    Same API, console and bill.

  • Rack-scaleBy request
    • GB300 NVL72
    • GB200 NVL72

    72 GPUs in one NVLink domain.On reserved terms.

  • us-east · fabric b
    Free 8Needs 16
    8 free < 16

    One fabric per job.

  • AccountacmeOne statement
    line
    cl_8f2k · GPU Clusters
    capacity
    spot ≤ 3.10
    cost center
    research-llm
    statement
    one · split by cost center
    metered
    per second

    Every product on one statement, split by cost center.

Checks every request against your policy, then places it on one fabric.
Fantasti OrchestratorBuilt to be run by agents
Capacity layer
Fabric aInfiniBand · us-east
  • GPU Clusters (highlighted)
    fx.clusters.create( gpu="H200:8", nodes=2)
    Private Preview
  • Workspaces
    fx.workspaces.create( gpu="H100:1")
    Early access
  • Sandboxes
    fx.sandboxes.create( ttl="15m")
    Early access
  • Flows (planned)
    @fantasti( gpu="H200:8", capacity="spot")
    Planned
  • Batch Inference
    fx.batch.create( input="s3://…")
    Private Preview
  • Serverless
    fx.jobs.create( image=…, gpu="H200:8")
    Early access
  • RequestPOST /v1/clustersREST
    { "name": "sft-01", "gpu": "H200:8", "nodes": 2, "capacity": { "mode": "spot", "max_price": "3.10" } }
    # the API answers{ "id": "cl_8f2k…", "status": "placing", "fabric": "single" }
  • GPUH200:8 × 2 (ok)
  • Fabricinfiniband (ok)
  • Regionus-* (ok)
  • Capacityspot (ok)
  • Max price≤ 3.10 (ok)
  • Your clustercl_8f2kIsolated
    HGX H200 8-GPU baseboard, line drawing8 GPUs per node
    nodes
    2 × H200:8
    fabric
    one · InfiniBand
    control
    your own plane
    state
    running

    Your own control plane anddedicated GPU nodes. No otheraccount runs on them.

    16 GPUs, dedicated
  • Rack-scaleBy request
    • GB300 NVL72
    • GB200 NVL72

    72 GPUs in one NVLink domain.On reserved terms.

  • us-east · fabric b
    Free 8Needs 16
    8 free < 16

    One fabric per job.

  • On the same layer
    • CPU Instances
    • Storage
    • Networking
    • Managed services

    Same API, console and bill.

  • AccountacmeOne statement
    line
    cl_8f2k · GPU Clusters
    capacity
    spot ≤ 3.10
    cost center
    research-llm
    statement
    one · split by cost center
    metered
    per second
Checks every request against your policy,then places it on one fabric.
  • ProductsOne API
    GPU Clusters
    Private Preview
    Workspaces
    Early access
    Sandboxes
    Early access
    Flows
    Planned
    Batch Inference
    Private Preview
    Serverless
    Early access
    fx.clusters.create( gpu="H200:8", nodes=2)
  • RequestPOST /v1/clustersREST
    { "name": "sft-01", "gpu": "H200:8", "nodes": 2, "capacity": { "mode": "spot", "max_price": "3.10" } }
  • Policy gates5 of 5
    GPU
    H200:8 × 2
    fabric
    infiniband
    region
    us-*
    capacity
    spot
    max price
    ≤ 3.10
  • Your clustercl_8f2kIsolated
    HGX H200 8-GPU baseboard, line drawing8 GPUs per node
    nodes
    2 × H200:8
    fabric
    one · InfiniBand
    control
    your own plane
    state
    running
    16 GPUs
  • us-east · fabric b
    Free 8Needs 16
    8 free < 16

    One fabric per job.

  • Rack-scaleBy request
    • GB300 NVL72
    • GB200 NVL72

    72 GPUs in one NVLink domain.On reserved terms.

  • On the same layer
    • CPU Instances
    • Storage
    • Networking
    • Managed services

    Same API, console and bill.

  • AccountacmeOne statement
    line
    cl_8f2k · GPU Clusters
    capacity
    spot ≤ 3.10
    cost center
    research-llm
    statement
    one · split by cost center
    metered
    per second
Fantasti OrchestratorCapacity layerFabric a · us-east
  • This request
  • Every other product
  • Planned
  • Passed over
§03 Capacity Reviewed 2026-10-10

The capacity layer comes from infrastructure providers. Fantasti builds the layer you use. You sign with Fantasti, call Fantasti and pay Fantasti.

  1. You One console, one API, one bill.

  2. Fantasti Builds the layer you use.

  3. The capacity layer NVIDIA GPU capacity, storage and networks.

§04 Compare

Pick the right product.

The Orchestrator is not another runtime. It places, limits and bills all of these.

Spot is a capacity mode, not a runtime. It applies to every column that says so.

TAB 02Pick the right product Source · Fantasti
Pick the right product
Question GPU Instances GPU Clusters Workspaces Early access Sandboxes Early access Flows Planned Batch Inference Serverless Early access
You bring An image, or your own stack A scheduler, or nothing but SSH Your editor and code An agent or test harness A Python flow A model and a dataset A container
Fantasti runs One machine with 1 or 8 GPUs Many 8-GPU nodes on one fabric A dev environment that scales to a cluster Short-lived isolated environments Each step, in order, on the GPU it names A sharded inference run One job or one endpoint
Ends When you stop it or its term ends When you stop it or its term ends When you stop it At its time-to-live When the last step does; then on its next trigger When every shard is done When the job exits or you stop the endpoint
Keeps Its volumes Its volumes and shared filesystem Your files Nothing; copy results out Every run, versioned Outputs and checkpoints Nothing on the container disk
Capacity On-demand, spot, reserved On-demand, spot workers, reserved On-demand head; workers on-demand or spot On-demand Per step: on-demand or spot Spot or on-demand On-demand or spot
Use it when You need a GPU and root The job spans nodes You are writing or debugging Code is untrusted or disposable The work repeats and must be reproducible The job is one model over one dataset The job is one container
Sheet
01 / 02
Title
Capacity and product comparison
Reviewed
2026-10-10
§05 API Preview API
Preview API · subject to change 
import fantasti

fx = fantasti.Client()                          # reads FANTASTI_API_KEY

cluster = fx.clusters.create(
    name="sft-01",
    gpu="H200:8",                               # 8-GPU nodes
    nodes=2,                                    # placed on one InfiniBand fabric
    capacity=fantasti.Spot(max_price=3.10),     # your ceiling, USD per GPU-hour
    interfaces=["kubernetes", "skypilot"],
)
cluster.wait("ready")
cluster.kubeconfig.save("~/.kube/fantasti-acme.yaml")
$ fantasti cluster create sft-01 --gpu H200:8 --nodes 2 --spot --max-price 3.10

Output · illustrative

request   H200:8 × 2 · one InfiniBand fabric · spot ≤ 3.10
pools     4 considered
placed    cl_8f2k… · us-east · fabric a
state     running · kubeconfig saved
{
  "mcpServers": {
    "fantasti": { "type": "http", "url": "https://mcp.fantasti.ai/mcp" }
  }
}

Tools · early access

catalog     gpus_list
sandboxes   sandboxes_create · sandboxes_exec · sandboxes_read_file · sandboxes_write_file · sandboxes_delete
workspaces  workspaces_list · workspaces_start · workspaces_stop
jobs        jobs_create · jobs_logs
usage       usage_get
curl -X POST https://api.fantasti.ai/v1/clusters \
  -H "Authorization: Bearer $FANTASTI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
        "name": "acme-train",
        "gpu": "H200:8",
        "nodes": 2,
        "capacity": { "mode": "spot", "max_price": "3.10", "currency": "USD", "unit": "gpu_hour" },
        "interfaces": ["kubernetes", "skypilot"]
      }'
# { "id": "cl_8f2k…", "status": "placing", "fabric": "single" }
# policy.yaml · applies to every request from team "research"
team: research
placement:
  gpus: [H100, H200, B200]
  fabric: infiniband            # multi-node jobs stay on one fabric
  regions: ["us-*"]
capacity:
  allow: [on_demand, spot, reserved]
  spot:
    max_price: { H100: 2.40, H200: 3.10 }   # USD per GPU-hour, your ceilings
limits:
  gpus: { H200: 32 }
  sandboxes: { concurrent: 200, max_ttl: 60m }
  spend: { monthly_usd: 40000 }
cost_center: research-llm

One package, one key.

Every product shares the client, the authentication and the error model. Every object reports where it was placed and what it costs.

Python
import fantasti
CLI
fantasti <noun> <verb>
REST
JSON over HTTPS
MCP
remote server

API referenceMCP server

§06 Controls

Set once, applied everywhere.

Policies written in the Orchestrator apply to every request, whichever product makes it and whoever makes it: a person, a pipeline or an agent.

Shared controls 5 controls · one path
One policy, every request. Schematic · not to scale.

A request reaches the Fantasti Orchestrator from a person, a pipeline or an agent, for any product: a person through CLI, here for GPU Clusters; a pipeline through Python, here for Batch Inference; an agent through MCP, here for Sandboxes. Every request passes the same five shared controls, in this order. 1, Identity: Single sign-on and role-based access. Set in console · access. 2, Placement policy: GPU, interconnect, region, capacity mode and max price. Set in policy.yaml · placement, capacity. 3, Quotas and limits: A pay-as-you-go quota that grows with you, and team limits your admins set. Set in policy.yaml · limits. 4, Cost centers: Every statement line is attributed. Set in policy.yaml · cost_center. 5, One statement: Every product, metered by the second. Set in statement · cost centers. Connections: A person to Fantasti Orchestrator; A pipeline to Fantasti Orchestrator: Every request; An agent to Fantasti Orchestrator; Identity to Placement policy; Placement policy to Quotas and limits; Quotas and limits to Cost centers; Cost centers to One statement.

Fantasti OrchestratorApplies to every request
  • CLIA personGPU Clusters
    $ fantasti cluster create sft-01 --gpu H200:8 --nodes 2 --spot --max-price 3.10
  • PythonA pipelineBatch Inference
    fx.batch.create(input="s3://…")# reads FANTASTI_API_KEY
  • MCPAn agentSandboxes
    sandboxes_create# OAuth, or a scoped key
  • 01Identity
    person
    SSO
    pipeline
    API key
    agent
    scoped key

    Single sign-on androle-based access.

  • 02Placement policy
    # policy.yamlplacement: gpus: [H100, H200, B200] fabric: infiniband regions: ["us-*"]capacity: allow: [on_demand, spot, reserved] spot: max_price: { H100: 2.40, H200: 3.10 }

    GPU, interconnect, region, capacity mode andmax price.

  • 03Quotas and limits
    # policy.yamllimits: gpus: { H200: 32 } spend: { monthly_usd: 40000 }

    A pay-as-you-go quota that grows withyou, and team limits your admins set.

  • 04Cost centers
    # policy.yamlcost_center: research-llm
    cluster
    cl_8f2k
    billed to
    research-llm

    Every statement line isattributed.

  • 05One statement
    metered
    per second
    split
    cost center

    Every product, meteredby the second.

Fantasti OrchestratorApplies to every request
  • A personGPU Clusters
    $ fantasti cluster create sft-01 --gpu H200:8 --nodes 2 --spot --max-price 3.10
  • A pipelineBatch Inference
    fx.batch.create(input="s3://…")# reads FANTASTI_API_KEY
  • An agentSandboxes
    sandboxes_create# OAuth, or a scoped key
  • 01Identity
    person
    SSO
    pipeline
    API key
    agent
    scoped key

    Single sign-on and role-basedaccess.

  • 02Placement policy
    # policy.yamlplacement: gpus: [H100, H200, B200] fabric: infiniband regions: ["us-*"]capacity: allow: [on_demand, spot, reserved] spot: max_price: { H100: 2.40, H200: 3.10 }

    GPU, interconnect, region, capacity mode and max price.

  • 03Quotas and limits
    # policy.yamllimits: gpus: { H200: 32 } spend: { monthly_usd: 40000 }

    A pay-as-you-go quota that grows withyou, and team limits your admins set.

  • 04Cost centers
    # policy.yamlcost_center: research-llm
    cluster
    cl_8f2k
    billed to
    research-llm

    Every statement line isattributed.

  • 05One statement
    metered
    per second
    split
    cost center

    Every product, metered bythe second.

  • Every request3 kinds
    A person
    CLI · GPU Clusters
    A pipeline
    Python · Batch Inference
    An agent
    MCP · Sandboxes
  • 01Identity
    person
    SSO
    pipeline
    API key
    agent
    scoped key

    Single sign-on and role-based access.

  • 02Placement policy
    # policy.yamlplacement: gpus: [H100, H200, B200] fabric: infiniband regions: ["us-*"]capacity: allow: [on_demand, spot, reserved] spot: max_price: { H100: 2.40, H200: 3.10 }

    GPU, interconnect, region, capacity mode and maxprice.

  • 03Quotas and limits
    # policy.yamllimits: gpus: { H200: 32 } spend: { monthly_usd: 40000 }

    A pay-as-you-go quota that grows with you, andteam limits your admins set.

  • 04Cost centers
    # policy.yamlcost_center: research-llm
    cluster
    cl_8f2k
    billed to
    research-llm

    Every statement line is attributed.

  • 05One statement
    metered
    per second
    split
    cost center

    Every product, metered by the second.

Fantasti Orchestrator
  • The path of every request
  • A request
One policy, every request. Schematic · not to scale.

A request reaches the Fantasti Orchestrator from a person, a pipeline or an agent, for any product: a person through CLI, here for GPU Clusters; a pipeline through Python, here for Batch Inference; an agent through MCP, here for Sandboxes. Every request passes the same five shared controls, in this order. 1, Identity: Single sign-on and role-based access. Set in console · access. 2, Placement policy: GPU, interconnect, region, capacity mode and max price. Set in policy.yaml · placement, capacity. 3, Quotas and limits: A pay-as-you-go quota that grows with you, and team limits your admins set. Set in policy.yaml · limits. 4, Cost centers: Every statement line is attributed. Set in policy.yaml · cost_center. 5, One statement: Every product, metered by the second. Set in statement · cost centers. Connections: Every request to Fantasti Orchestrator; Identity to Placement policy; Placement policy to Quotas and limits; Quotas and limits to Cost centers; Cost centers to One statement.

Fantasti OrchestratorApplies to every request
  • Every request3 kinds
    A person
    CLI · GPU Clusters
    A pipeline
    Python · Batch Inference
    An agent
    MCP · Sandboxes
  • 01Identity
    person
    SSO
    pipeline
    API key
    agent
    scoped key

    Single sign-on and role-basedaccess.

  • 02Placement policy
    # policy.yamlplacement: gpus: [H100, H200, B200] fabric: infiniband regions: ["us-*"]capacity: allow: [on_demand, spot, reserved] spot: max_price: { H100: 2.40, H200: 3.10 }

    GPU, interconnect, region, capacity mode and maxprice.

  • 03Quotas and limits
    # policy.yamllimits: gpus: { H200: 32 } spend: { monthly_usd: 40000 }

    A pay-as-you-go quota that grows with you,and team limits your admins set.

  • 04Cost centers
    # policy.yamlcost_center: research-llm
    cluster
    cl_8f2k
    billed to
    research-llm

    Every statement line is attributed.

  • 05One statement
    metered
    per second
    split
    cost center

    Every product, metered by the second.

  • The path of every request
  • A request
§07 Agents Early access

Built to be run by agents.
Under written rules.

The platform's volume work is built for AI agents: answering, diagnosing, placing and pricing. Each has one job and a limit on what it may decide. High-stakes decisions pass a second, senior review. The owner sets the rules.

Agents
13 roles · 3 planned
Decide
act · ask · hand off
Levels
desk · senior review · the owner

How the agents work

Operating model. Source · Fantasti. Reviewed 2026-10-10.

Raise a quota: the Quota analyst reads your payment history and commitment, the risk rules check the increase, and above the limits of the Direct plan it goes to senior review.

Three levels. 01 Desk: front-line and technical agents in 10 teams (Sales, Support, Engineering, Security, Review, Legal, Monitoring, Risk, Capacity, Release) decide everything inside their written playbook and limits. 02 Senior review, by the orchestrator tier, takes the high-stakes decisions: Refunds and credits above $500; Suspending or terminating an account; Reservations and capacity requests to the capacity layer; Security disclosures; Exceptions to a playbook. The reviewer is a different agent from the one that proposed the decision, and it records its reasoning. Everything else, agents decide inside their playbooks and log. 03 The owner: The rules and limits themselves, and the few acts only a legal person can perform. Signs contracts and legal documents for Fantasti. Owns the bank and payment accounts and passes their identity checks. Answers legal process and regulators, and files taxes. Sets and changes the hard limits and the decision rights. The levels work under the written rules, set by the owner: Catalog and prices (what a customer pays), Your policy (quotas, caps, ceilings), Risk rule set (allow, verify, hold, block), Playbooks (versioned and reviewed). Every decision is logged in a decision record: role · playbook · tier · inputs · rule · reasoning · outcome · review. Connections: Desk to Senior review: Hand off; Senior review to The owner: Digest; The owner to Written rules: Sets; Written rules to Decision record: Writes.

Written rulesSet by the owner
  • 01Desk10 teams

    Everything inside their written playbook and limits.

    • Sales
    • Support
    • Engineering
    • Security
    • Review
    • Legal
    • Monitoring
    • Risk
    • Capacity
    • Release
    Anouk
    Front desk · AI agent
    Tamsin
    Support desk · AI agent
    Kofi
    QA reviewer · AI agent
    Hiro
    Quota analyst · AI agent
    Rafael
    Risk and fraud officer · AI agent
    Rafael
    Risk and fraud officer · AI agent
    Kofi
    QA reviewer · AI agent
    Freya
    Release manager · AI agent
  • 02Senior reviewOrchestrator tier
    • Refunds above $500
    • Account suspension
    • Capacity requests
    • Security disclosures
    • Playbook exceptions

    The reviewer is a different agent from the one that proposed the decision, and it recordsits reasoning.

    None
    decided at the desk
    Ingrid
    Head of Review · AI agent
    Ingrid
    Head of Review · AI agent
    Ingrid
    Head of Review · AI agent
  • 03The ownerRules and limits

    The rules and limits themselves, and the few acts only a legal person can perform.

    • Contracts and legal documents
    • Bank and payment accounts
    • Legal process and regulators
    • Hard limits and decision rights
  • Catalog and prices

    What a customer pays

  • Your policy

    Quotas, caps, ceilings

  • Risk rule set

    Allow, verify, hold, block

  • Playbooks

    Versioned and reviewed

  • RecordDecision record8 fields · one per decision
    role
    front-desk, support-desk, qa-reviewer
    role
    quota-analyst, risk-officer
    role
    risk-officer
    role
    qa-reviewer, release-manager
    rule
    catalog and prices, playbooks
    rule
    your policy, risk rule set
    rule
    risk rule set
    rule
    playbooks
    tier
    front-line, senior
    tier
    senior
    tier
    senior
    tier
    senior, orchestrator-tier
    review
    none
    review
    senior review
    review
    senior review
    review
    senior review

    Every decision is logged: role · playbook · tier · inputs · rule · reasoning · outcome ·review.

  • 01Desk10 teams

    Everything inside their written playbookand limits.

    • Sales
    • Support
    • Engineering
    • Security
    • Review
    • Legal
    • Monitoring
    • Risk
    • Capacity
    • Release
    Anouk
    AI agent
    Front desk
    Tamsin
    AI agent
    Support desk
    Kofi
    AI agent
    QA reviewer
    Hiro
    AI agent
    Quota analyst
    Rafael
    AI agent
    Risk and fraud officer
    Rafael
    AI agent
    Risk and fraud officer
    Kofi
    AI agent
    QA reviewer
    Freya
    AI agent
    Release manager
  • 02Senior review
    • Refunds above $500
    • Account suspension
    • Capacity requests
    • Security disclosures
    • Playbook exceptions

    The reviewer is a different agent from theone that proposed the decision, and itrecords its reasoning.

    None
    decided at the desk
    Ingrid
    Head of Review · AI agent
    Ingrid
    Head of Review · AI agent
    Ingrid
    Head of Review · AI agent
  • 03The ownerRules and limits

    The rules and limits themselves, and thefew acts only a legal person can perform.

    • Contracts and legal documents
    • Bank and payment accounts
    • Legal process and regulators
    • Hard limits and decision rights
  • Written rulesSet by the owner
    Catalog and prices
    Your policy
    Risk rule set
    Playbooks
  • RecordDecision record8 fields
    role
    front-desk, support-desk,
    qa-reviewer
    role
    quota-analyst, risk-officer
    role
    risk-officer
    role
    qa-reviewer, release-manager
    rule
    catalog and prices, playbooks
    rule
    your policy, risk rule set
    rule
    risk rule set
    rule
    playbooks
    tier
    front-line, senior
    tier
    senior
    tier
    senior
    tier
    senior, orchestrator-tier
    review
    none
    review
    senior review
    review
    senior review
    review
    senior review

    Every decision is logged: role · playbook ·tier · inputs · rule · reasoning · outcome· review.

  • The traced decision
  • Works under the rules
  • Digest to the owner
  • Goes to senior review
§08 Stages Reviewed 2026-10-10

Every product carries a stage.

Release stages, defined

Sheet
02 / 02
Title
Release stages
Reviewed
2026-10-10
§09 Next Private Preview

Start with the product you need first. The others are on the same account.

Request access

Next