Last Updated: March 2026
1. Scope of This SLA
This Service Level Agreement (“SLA”) covers the TrueFoundry-managed components of the AI Gateway across two deployment models:
1.1 SaaS Deployment
In the SaaS model, TrueFoundry hosts and operates the control plane. This SLA covers:
Gateway control-plane APIs hosted by TrueFoundry
Routing, governance, and policy-enforcement services
Observability, analytics, and management dashboards
Prompt registry, guardrails engine, and MCP control surfaces
1.2 On-Premises Deployment
In the on-premises model, TrueFoundry deploys and manages specific components within the customer’s environment. This SLA covers:
TrueFoundry-managed gateway control-plane services running in the customer’s infrastructure
Routing, governance, and observability interfaces managed by TrueFoundry
Prompt, guardrails, and MCP control surfaces deployed and maintained by TrueFoundry
1.3 Exclusions
This SLA does NOT apply to:
Customer-managed infrastructure, networking, compute, or storage resources
Upstream LLM or tool providers (OpenAI, Anthropic, AWS Bedrock, Google, etc.)
Customer application code or custom integrations built on top of the gateway
Data-plane components processing inference traffic within the customer’s network (on-prem)
Third-party services or connectors not operated by TrueFoundry
2. Definitions
| Term |
Definition |
|
Service
|
The TrueFoundry-managed AI Gateway control-plane components, whether hosted by TrueFoundry (SaaS) or deployed on-premises
|
|
Applicable Month
|
A calendar month measured in UTC
|
|
Downtime
|
A minute during which the Service is unavailable or materially degraded as measured by TrueFoundry monitoring
|
|
Scheduled Maintenance
|
Planned maintenance announced at least 72 hours in advance
|
|
Monthly Uptime Percentage
|
Availability calculated per Section 4
|
|
Service Credit
|
A percentage credit applied to future invoices (Enterprise customers only)
|
|
Service Fees
|
Monthly fees paid for the AI Gateway Service
|
3. Service Availability Commitment
3.1 Uptime Target
This SLA’s uptime commitment and Service Credits apply only to Enterprise customers. All other plans receive best-effort availability without a formal uptime guarantee.
| Plan |
Deployment Model |
Monthly Uptime Commitment |
|
Developer
|
SaaS
|
Best effort (no SLA)
|
|
Pro
|
SaaS
|
Best effort (no SLA)
|
|
Pro Plus
|
SaaS
|
99.9%
|
|
Enterprise
|
SaaS
|
99.99%
|
|
Enterprise
|
On-Premises
|
99.99%
|
Note: 99.9% uptime allows approximately 43 minutes of Downtime per 30-day month.
3.2 Deployment-Specific Considerations
SaaS: TrueFoundry is fully responsible for the availability of the hosted control plane, including the underlying infrastructure, networking, and platform operations.
On-Premises: TrueFoundry is responsible for the availability of the managed software components it deploys and maintains. Availability may be affected by factors outside TrueFoundry’s control (customer infrastructure, network, hardware). Such factors are excluded from Downtime calculations (see Section 9).
4. Uptime Measurement and Calculation
4.1 What Counts as Downtime
A minute is counted as Downtime if 5% or more of valid API requests to TrueFoundry-managed control-plane endpoints return HTTP 5xx errors attributable to TrueFoundry components.
4.2 Monthly Uptime Formula
Monthly Uptime % = ((Total Minutes in Month − Downtime Minutes) ÷ Total Minutes in Month) × 100
5. Scheduled Maintenance
Standard maintenance: at least 72 hours advance notice
Major upgrades: at least 7 calendar days advance notice
Preferred window: low-traffic hours coordinated with the customer
Designed for zero-downtime rolling updates whenever possible
Scheduled Maintenance does not count toward Downtime.
On-Premises note: Maintenance windows for on-prem deployments will be coordinated with the customer’s change-management process.
6. Emergency Maintenance
TrueFoundry may perform emergency maintenance to address security vulnerabilities, data-integrity risks, or platform-wide incidents. Emergency maintenance counts toward Downtime only if it causes more than 30 minutes of actual unavailability.
7. Service Credits
7.1 Eligibility
Service Credits apply only to Enterprise customers with active paid subscriptions, regardlessof deployment model (SaaS or On-Premises). Developer, Pro, and Pro Plus plans are noteligible for Service Credits.
A Service Credit is a monetary credit applied to a future invoice, calculated as a percentage ofthe monthly fees paid for the TrueFoundry AI Gateway for the affected calendar month.
7.2 Service Credit Table (For Enterprise Tier Only)
| Monthly Uptime Percentage
|
Service Credit |
|
≥99.0% but <99.99%
|
1%
|
|
≥95.0% but <99.0%
|
2%
|
|
<95.0%
|
5%
|
7.3 Credit Terms
Credits are applied to future invoices only and are not redeemable for cash
Credits are non-transferable between accounts or services
Credits are capped at 100% of the monthly AI Gateway Service Fees
7.4 Credit Request Process
- Submit a request within 45 days of the end of the affected month
- Include: Account Name, Impacted Month, Deployment Model (SaaS or On-Prem), and Incident Description
- Submit via the support portal or email
support@truefoundry.com - Credits will be applied to the next applicable invoice (no cash refunds)
8. Support and Incident Response
8.1 Incident Severity Levels
| Severity
|
Description |
|
P1
|
Complete service outage or active security breach affecting production
|
|
P2
|
Major feature unavailable with no reasonable workaround
|
|
P3
|
Partial degradation with a viable workaround available
|
|
P4
|
Minor issue, cosmetic defect, or general inquiry
|
8.2 Response Targets (Enterprise Only)
The following response targets apply to Enterprise customers across both SaaS and On-Premises deployments:
| Severity |
Enterprise |
Pro Plus |
|
P1 — Critical
|
4 hours
|
8 hours
|
|
P2 — High
|
8 hours
|
12 hours
|
|
P3 — Medium
|
2 business days
|
4 business day
|
|
P4 — Low
|
4 business days
|
8 business days
|
Response targets represent the time to initial acknowledgement and triage, not resolution.
On-Premises note: For on-prem incidents, TrueFoundry may require remote access or customer assistance to diagnose issues within the customer’s environment. Response targets begin once TrueFoundry has been granted the necessary access.
9. Exclusions
Downtime does not include, and no Service Credits will be issued for, issues caused by:
Upstream LLM or tool provider outages or degradation
Customer misconfiguration, misuse, or unauthorized modifications to managed components
Force majeure events (natural disasters, war, government actions, etc.)
Customer network, hardware, or infrastructure failures (especially relevant for on-prem deployments)
Beta, preview, or experimental features
Usage exceeding documented capacity limits without prior arrangement
Failure by the customer to maintain the environment prerequisites specified by TrueFoundry (on-prem)
Customer refusal or delay in granting TrueFoundry the access needed to perform maintenance or incident response (on-prem)
10. Data & Security
All data encrypted in transit (TLS 1.2+) and at rest
SOC 2 Type II compliant
GDPR compliant
RBAC, SSO, and audit logging (tier-dependent feature availability)
TrueFoundry does not use customer data for model training
11. Governing Terms
This SLA is incorporated into and governed by the Master Subscription Agreement (MSA)
Service Credits are the sole and exclusive remedy for failure to meet the uptime commitment
TrueFoundry may update this SLA with 30 days prior written notice
In the event of conflict between this SLA and the MSA, the MSA governs
12. Contact Information
