Support Early access
Every case gets a severity,
an owner and a clock.
Fantasti support comes in five plans. It is set up so that an AI agent takes each case as it arrives, at any hour, and says it is an AI. On paid plans an escalation goes to senior review in the business days your plan states.
gpu-node-02 reports Xid 79, job is degraded
cl_8f2k… · H200:8 × 2 · Direct
- opened in
- Slack · #fantasti-acme
- taken by
- Tamsin Support deskAI agent evidence attached
- owner
- Priya Infrastructure deskAI agent
- state
- with infrastructure desk
- First response target, any hour
- 5minutes
- Senior review of an escalation
- Same business day
One support case, as an example: opened in a shared Slack channel on the Direct plan at SEV 2. In this example it is taken by Tamsin · Support desk · AI agent with the evidence attached, and owned by Priya · Infrastructure desk · AI agent. The targets shown are those of this plan and severity.
Support at a glance
5 plans
- 01 Basic Included . First response target, any hour: 15 minutes
- 02 Developer $29 per month. First response target, any hour: 10 minutes
- 03 Team $100 per month. First response target, any hour: 5 minutes
- 04 Direct $1,000 per month. First response target, any hour: 5 minutes
- 05 Enterprise Custom . First response target, any hour: In your order form
Price per month, per organizationFirst response target, any hour
Source · Fantasti support terms
- Severities
- 3
- You set the severity. If we change it, we tell you why.
- Channels
- Console · Email · Slack · API
- One case record, whichever way you open it.
- Coverage
- AI agents at any hour
- On every plan. Escalations go to senior review in business days, as your plan states.
Five plans. Prices on the page.
A flat fee per month, per organization. No plan takes a percentage of your spend.
Billed on your Fantasti statement with your usage. Change plans at any time: an upgrade applies at once, a downgrade at your next billing date.
- A plan's place among the five.
- One answer for the plans under the bracket.
- A plan that does not include the feature.
| Feature | 01 Basic | 02 Developer | 03 Team | 04 Direct | 05 Enterprise |
|---|---|---|---|---|---|
| Price per month, per organization | Basic Included | Developer $29 | Team $100 | Direct $1,000 | Enterprise Custom |
| For | Basic Everyone | Developer Individual developers | Team Startups running production | Direct Teams with reserved or multi-node capacity | Enterprise Large organisations Talk to the solutions desk |
| Who answers | All five plans AI agents, at any hour. Each one says it is an AI in its first message and in its signature. | ||||
| First response target, any hour | Basic 15minutes | Developer 10minutes | Team and Direct 5minutes | Enterprise Custom, in your order form | |
| Queue | Basic and Developer Standard | Team to Enterprise Priority | |||
| Senior review of an escalation | Basic Billing disputes and security reports only | Developer Within 2 business days | Team Within 1 business day | Direct Same business day. SEV 1 goes straight to senior review at any hour. | Enterprise Custom, in your order form |
| Workload debugging | Basic Not included | Developer to Enterprise By the technical-support tier of AI agents, from logs and metrics read with your consent | |||
| Architecture review | Basic and Developer Not included | Team to Enterprise One per quarter, by the architect, an AI agent | |||
| Channels | Basic to Team Console and email | Direct Console, email and a shared Slack channel | Enterprise Console, email and private Slack Connect | ||
| Account manager | Basic to Team Not included | Direct Named account manager, an AI agent | Enterprise Named account manager and a named account lead, both AI agents | ||
| Infrastructure desk | Basic to Team Not included | Direct Infrastructure desk of AI agents: proactive monitoring of your clusters, capacity planning and reservation handling | Enterprise Dedicated infrastructure desk of AI agents | ||
| Automatic quota growth | Basic and Developer Included | Team to Enterprise Included, with fast-track | |||
| Incident reviews | Basic to Direct Not included | Enterprise Written, for your incidents | |||
A platform fault is not a paid feature. When something breaks on our side, we work it at SEV 1 for every affected account on every plan.
Senior review is a second AI agent, on the orchestrator tier. It is never the agent that worked the case, and it records its reasoning.
Three severities, defined by impact.
Pick the severity that matches the effect on your work. If the effect changes, change the severity.
| Severity | Definition | Examples |
|---|---|---|
| SEV 1 · Critical | Production is down, or data is at risk. |
|
| SEV 2 · High | Production is degraded. |
|
| SEV 3 · Normal | A question, or an issue outside production. |
|
SEV 1 is worked around the clock and needs a contact on your side who stays reachable. If we cannot reach you, we lower the case to SEV 2 and tell you.
Two clocks: minutes for a first reply, business days for senior review.
Two clocks. First response runs from the moment a case is opened to an AI agent's first reply: a target of minutes, at any hour. Senior review runs from an escalation to a reply from a second, more senior AI agent: a target of business days. The work that follows, to resolution, is not measured by either target. Connections: A case is opened to An AI agent replies; An AI agent replies to Work to resolution; A case is escalated to Senior review replies.
The first-response clock starts.
The first-response clock stops.
Not measured by a target.
The review clock starts.
The review clock stops.
The first-response clock starts.
The first-response clock stops.
Not measured by a target.
The review clock starts.
The review clock stops.
The first-response clock starts.
The first-response clock stops.
Not measured by a target.
The review clock starts.
The review clock stops.
- Measured by the target
- Not measured
| Severity | 01Basic | 02Developer | 03Team | 04Direct | 05Enterprise |
|---|---|---|---|---|---|
| First response targets, AI agentMinutes, at any hour | |||||
| SEV 1, SEV 2, SEV 3 | Basic 15minutes | Developer 10minutes | Team 5minutes | Direct 5minutes | Enterprise Custom, in your order form |
| Senior review of an escalationBusiness days | |||||
| SEV 1 | Basic Billing disputes and security reports only | Developer Within 2 business days | Team Within 1 business day | Direct Same business day. Straight to senior review at any hour | Enterprise Custom, in your order form |
| SEV 2, SEV 3 | Basic Billing disputes and security reports only | Developer Within 2 business days | Team Within 1 business day | Direct Same business day | Enterprise Custom, in your order form |
They are targets for response, not a service level, and they carry no service credits. Service availability commitments are in your contract and the SLA.
- Sheet
- 01 / 02
- Title
- Support plans
- Unit
- USD per month
- Reviewed
- 2026-10-09
An AI agent takes the case.
An escalation gets a senior review.
The agent reads the state of your account before it replies: quotas, placement records, node health, spot stops, invoices.
A case can be opened from Console, Email, Slack, API. Every channel writes to one case record. The desk is set up so that an AI agent takes every case as it arrives, at any hour, and says it is an AI. It reads the state of the account (quotas, placement records, node health, spot stops, invoices), then fixes the problem and says what it changed, or escalates the case with the evidence attached. An escalation goes to senior review, a second AI agent on the orchestrator tier, within the business days your plan states. On Direct and Enterprise the infrastructure desk, also AI agents, owns the case and watches the cluster. If the fault is in the infrastructure underneath, we work it with the capacity layer and keep your case updated. Connections: Console to case_7c2m…; Email to case_7c2m…; Slack to case_7c2m…; API to case_7c2m…; case_7c2m… to AI agent; AI agent to case_7c2m…: Fixes it; AI agent to Senior review: Escalates; AI agent to Infrastructure desk: Owns it; Infrastructure desk to Capacity layer.
A severity,an ownerand a clock.
Takes every case as it arrives and saysit is an AI. Reads the account first:
- placement records
- invoices
- quotas
- node health
- spot stops
Fixes it and says what changed, orescalates.
A second AI agent, on the orchestrator tier. It gets the casewith the evidence attached.
On Direct and Enterprise. Ownsthe case and watches the cluster.
For a fault in theinfrastructureunderneath.
A severity, an ownerand a clock.
Takes every case as it arrivesand says it is an AI. Reads theaccount first:
- placement records
- invoices
- quotas
- node health
- spot stops
Fixes it and says what changed,or escalates.
A second AI agent, on theorchestrator tier. It gets thecase with the evidence attached.
On Direct and Enterprise.Also AI agents. Owns thecase and watches thecluster.
For a fault in theinfrastructureunderneath.
A severity,an ownerand a clock.
Takes every case as it arrives and saysit is an AI. Reads the account first:
- placement records
- invoices
- quotas
- node health
- spot stops
Fixes it and says what changed, orescalates.
A second AI agent, on the orchestratortier. It gets the case with theevidence attached.
On Direct and Enterprise. Owns the caseand watches the cluster.
For a fault in the infrastructureunderneath.
- Every case
- As the case needs
import fantasti
fx = fantasti.Client()
case = fx.support.cases.create(
severity="sev2",
resource="cl_8f2k",
summary="gpu-node-02 reports Xid 79, job is degraded",
attach=["placement", "node_events"],
)
print(case.id, case.state, case.target)Output · illustrative
case_7c2m… with_agent first response · 5 minutes{
"mcpServers": {
"fantasti": { "type": "http", "url": "https://mcp.fantasti.ai/mcp" }
}
}Support tools · planned
docs_search
support_open_ticket
support_get_ticketcurl -X POST https://api.fantasti.ai/v1/support/cases \
-H "Authorization: Bearer $FANTASTI_API_KEY" \
-H "Content-Type: application/json" \
-d '{ "severity": "sev2", "resource": "cl_8f2k",
"summary": "gpu-node-02 reports Xid 79, job is degraded",
"attach": ["placement", "node_events"] }'
# { "id": "case_7c2m…", "state": "with_agent" }If the agent can fix the problem it does, and it tells you what it changed. If it cannot, it escalates the case with the evidence attached, so nobody asks you to explain your environment twice.
- EscalationAI agent, then senior review for an escalation, then the capacity layer for a fault underneath.
- Hands-on workHardware and data-center issues go to the operations team of the provider that runs the hardware, through the infrastructure desk.
- One recordConsole, email, Slack and API write to the same case.
- Open from codeCreate and follow cases through the API or the Python SDK.
Support is designed to be operated end to end by AI agents, under written rules and hard limits. Every decision is logged.
Support in the channel your team already has open.
-
YouCustomer
gpu-node-02 reports Xid 79, job is degraded.
- id
- case_7c2m…
- opened in
- Slack · #fantasti-acme
-
Tamsin Support deskAI agent
I am Tamsin, an AI agent on the Fantasti support team.
This is now a case at SEV 2 on your Direct plan. I attached the placement record and the node events for cl_8f2k.
Signature: Tamsin · Support desk · AI agent
- severity
- SEV 2 · High
- taken by
- AI agent · evidence attached
- attach
- placement, node_events
- target
- first response · 5 minutes
-
Priya Infrastructure deskAI agent
I am Priya, an AI agent on the Fantasti engineering team.
I own this case from here. I am reading the node events for gpu-node-02.
Signature: Priya · Infrastructure desk · AI agent
- owner
- infrastructure desk · AI agent
- state
- with infrastructure desk
AI agentEvery agent identifies itself as an AI in its first message and in its signature.
An example. A customer posts a problem in a shared channel. Tamsin · Support desk · AI agent opens a case with a severity and attaches the evidence. Priya · Infrastructure desk · AI agent takes ownership. Each message writes to the same case record.
Named roles for the accounts that need them.
- 01 Included in: DirectEnterprise
Lucia
Sales deskAI agent
Account manager
A named AI agent that knows your contract, your reservations and your plans, and follows every case of yours that goes to senior review.
Reports toIdris · Head of Sales · AI agent
Tasks
- Answers product and pricing questions from the catalog
- Reads every access request and reviews it for fit
- Drafts quotes for on-demand and reserved capacity
- Is the named account manager on Direct and Enterprise
Reports it files
- Desk report · daily
- Access review note · per request
- 02 Included in: DirectEnterprise
Priya
Infrastructure deskAI agent
Infrastructure desk
AI agents that watch your clusters, plan capacity with you and handle your reservations. Dedicated to one organization on Enterprise.
Reports toSoren · Head of Engineering · AI agent
Tasks
- Monitors the clusters of Direct and Enterprise accounts
- Plans capacity with the account and handles its reservations
- Takes the case that needs work on a node or a cluster
- Reports an infrastructure fault and follows it to repair
Reports it files
- Cluster health report · weekly, per account
- Fault report · per fault
- 03 Included in: Enterprise
Idris
Head of SalesAI agent
Account lead
A named AI agent that answers for the account.
Reports toMaren · Orchestrator · AI agent
Tasks
- Leads the sales desk and the solutions desk
- Carries an enterprise request from the first message to a contract that is ready for signature
- Keeps enterprise intake inside what can be signed, and opens the waitlist when it is not
- Answers for each Enterprise account as its named account lead
Reports it files
- Sales report · weekly
- Contract file · per contract
- Enterprise queue · weekly
Every role here is an AI agent, and each one says so. Enterprise adds a named account lead. The avatar is a two-letter monogram. No agent has a photograph or a drawn face.
The desks behind every plan
-
Support desk
Every plan
AI agents that answer billing and technical questions at any hour, by console and email.
-
Technical-support tier
Developer to Enterprise
AI agents that debug a workload from its logs and metrics, read with your consent.
-
Architect
Team to Enterprise
An AI agent that reviews your architecture once per quarter and writes up what it would change.
Your quota grows with you. Up to the limits of the Direct plan there is no sales step. The Quota analyst raises your quota in steps and writes a decision record.
- BasisYour quota grows automatically, on your payment history and your commitment.
- Larger stepsAbove those limits, the Quota analyst prepares the case and senior review decides.
- Fast-trackOn Team and above, quota reviews run ahead of the queue.
- RefusalsA refusal or a reduction always goes to senior review.
- Applies toA quota is a control on pay-as-you-go accounts. Prepaid credit has no instance quota: it is limited by its balance and by available capacity. Enterprise quotas are set by contract.
- Who reviews Hiro Quota analystAI agent : identity card
- At review
- quota 16peak 13headroom 3
- Now
- quota 24peak 18headroom 6
Decisiondec_4k9t…
Hiro Quota analystAI agent
raised · 16 → 24desk · inside the limits
Illustrative. Demand for H200 GPUs on the account acme rises in steps toward its quota of 16. A dashed review line sits under the quota. When demand reaches the review line, at a peak of 13, the quota is raised one step to 24 and a decision record is written. Demand passes the old quota later and never reaches a ceiling.
What support covers.
Covered
- Fantasti products in Private Preview, Public Preview and General Availability.
- Early access features, in the preview's own channel.
- Billing, account, quota and access questions on every plan.
- Workload debugging on Developer and above, from logs and metrics read with your consent. Kubernetes, Ray and SkyPilot are included, as they run on a Fantasti cluster.
Not covered
- Your model or application code itself.
- Third-party software you install, beyond configuration guidance.
- Workload performance tuning, except as an Enterprise engagement.
- Capacity guarantees. Reserve capacity to guarantee it.
Questions about support.
Is technical support included without a paid plan?
Yes. Basic is included with every account: the desk is designed so that AI agents answer billing and technical questions at any hour, with a first response target of 15 minutes. Paid plans add workload debugging, shorter targets, more channels and senior review of escalations.
Who answers?
By design, an AI agent, and it says so in its first message and in its signature. An escalation goes to senior review, a second AI agent on the orchestrator tier, within the business days your plan states. On Basic, senior review covers billing disputes and security reports.
Can I talk to a person?
Until the desks open with the first cohort, the owner reads every request and answers it. After that every desk is AI agents, and each one says so in its first message and in its signature. Ask on any channel to escalate. The case goes to senior review, a second AI agent on the orchestrator tier, with the notes of the agent that worked it, so you do not repeat yourself.
The fault is in the infrastructure underneath. Who do I contact?
Us. You open one case with Fantasti. If the fault is in the infrastructure underneath, we work it with the capacity layer and keep your case updated.
Hardware and data-center issues go to the operations team of the provider that runs the hardware, through the infrastructure desk.
What happens during a platform incident?
We work it at SEV 1 for every affected account, whatever its plan. You do not need a case for an incident we have already announced unless your impact is different.
Can I change plans for a launch?
Yes. Upgrade before the event and the new targets apply from that moment. Downgrade at your next billing date.
Are response targets an SLA?
They are targets for response, not a service level, and they carry no service credits. Service availability commitments are in your contract and the SLA.
Which languages?
English. Portuguese by request.
Do previews get support?
Private Preview and Public Preview features are covered by your plan. Early access features are supported in the preview's own channel, without response targets.
- Sheet
- 02 / 02
- Title
- Support terms
- Reviewed
- 2026-10-09