Terms and Conditions

Last Updated: January 7, 2026

Introduction

This document outlines the Service Terms and Service Level Agreements (SLA) governing your use of cloudidr's services. cloudidr offers multiple distinct services: Disaster Recovery Compute Service (DRCS), Schedule Compute Service (SCS) / FlexCompute, GPU Service, Hosted Models, and LLM Ops, each with specific terms and SLAs.

These Service Terms and Service Level Agreement ('Service Terms') supplement and amend the Customer Agreement between CloudIDR and Customer. In the event of any conflict or inconsistency between these Service Terms and the Customer Agreement, these Service Terms shall take precedence and control with respect to the Services described herein. For matters not explicitly addressed in these Service Terms, the Customer Agreement shall govern. Terms defined in these Service Terms supersede any conflicting definitions in the Customer Agreement.

1. Disaster Recovery Compute Service (DRCS)

1.1 Service Description

DRCS provides high-availability compute services enabling businesses to recover their applications in a designated failover region. Customers can purchase one or more Disaster Recovery (DR) policies for their failover regions. DRCS is specifically designed and must be used exclusively for DR events.

1.2 Disaster Recovery Policy Terms

Each DR policy includes:

• Designated region and availability zone (AZ)

• Specified instance types and quantities

• 12-month term commitment

• Usage limitations:

  • 96 total hours of usage per instance within 12 months

  • Maximum 48 continuous hours of usage per instance

  • Concurrent utilization of all policy-defined instances

  • Maximum 3 activation requests per 12-month period

• Policy terms:

  • Non-refundable

  • Modifications permitted only for same-region upgrades, subject to CloudIDR's capacity

  • All instances within a policy must be utilized in the same Region/AZ

• Instance substitution:

  • In rare cases when exact instance type is unavailable, cloudidr may substitute with an equivalent instance of similar or higher family/size

  • Substitution maintains or exceeds specifications (CPU, memory, network performance)

  • Customer will be notified of substitution

  • No additional charges for higher-tier substitutions

1.3 DR Policy Availability (DRPA)

• Target availability: 99.999% for policy-specified instances

• Measurement: Based on instance availability during DR event activations

• Service Credits eligibility for non-compliance with SLA

1.4 DRCS Service Level Agreement

1.4.1 Availability Target

• 99.999% instance availability (excluding scheduled maintenance)

• Maximum allowable downtime: 6 minutes per year during policy usage

1.4.2 Availability Exceptions

• Force majeure events

• Cloud service provider failures in their AZ or regions

• Scheduled maintenance (with 48-hour advance notice)

1.4.3 Service Credits

• Below 99.999%: 10% credit

• Below 99.0%: 25% credit

• Non-availability: 100% credit for affected instance-minutes

2. Schedule Compute Service (SCS) / FlexCompute

2.1 Service Description

SCS (also known as FlexCompute) provides discounted AWS EC2 compute resources (GPU and CPU) at discount compared to AWS On-Demand pricing. The service operates in two modes:

• Advance capacity reservation for customers with flexible usage schedules

• Billing based on usage-minutes, rounded up to the nearest minute

• Requires advance booking

• Instances provisioned directly in customer's AWS account

• Pricing available at cloudidr.com/pricing

2.2 Service Terms

2.2.1 Scheduled Reservations (SCS) Terms:

• Refund Policy:

  • Full refund: Within 10 minutes of creation and before policy start

  • 80% credit: When cancelled ≥24 hours before start time

  • Early termination: Unused instance-hours credited to maximum of 20% of purchase price

• Modifications: Not permitted after booking

• Billing: Pre-paid based on reserved capacity and credited at the end of the moth

2.3 Service Level Agreement

2.3.1 Scheduled Reservations (SCS) SLA:

• Availability Target: 99% during reservation period

• Maximum Downtime: Approximately 14 minutes per 24/7 usage

• Service Credits: 100% credit for unavailable reserved instance-minutes

• Instance Availability: Best-effort basis; cloudidr does not guarantee 100% availability

2.4 Customer Responsibilities

• Maintain valid AWS account with appropriate permissions

• Provide cloudidr with necessary IAM roles for instance provisioning

• Monitor and manage instances through AWS Console or Cloudidr dashboard

• Terminate instances when no longer needed to stop billing (Pay-As-You-Go)

• Utilize scheduled reservations during booked time windows (Scheduled Reservations)

3. GPU Service

3.1 Service Description

GPU Service provides SSH-only access to discounted AWS GPU instances. Customers receive shell access to GPU capacity at discount compared to AWS On-Demand pricing.

3.2 Service Terms

• Access Method: SSH access only; no AWS Console access to these specific instances

• Billing: Usage billing based on actual minutes of usage 

• Instance Availability: Best-effort basis; no guaranteed availability

• Data Location: Instances run in AWS regions

• Session Management: Sessions may be terminated with notice if capacity reallocation is needed

• Cancellation and refunds: similar to schedule compute service (SCS).

3.3 GPU Service Level Agreement

• No SLA commitment; service provided on best-effort basis

• Cloudidr will provide reasonable notice (minimum 24 hours when possible) before session termination due to capacity needs

• No service credits available

3.4 Customer Responsibilities

• Maintain secure SSH keys and access credentials

• Back up data regularly; cloudidr is not responsible for data loss

• Comply with SSH access policies and security best practices

• Terminate sessions when not in use to minimize costs

4. Hosted Models

4.1 Service Description

Hosted Models provides managed LLM hosting where Cloudidr installs and operates AI models (including OpenAI, Mistral, DeepSeek, and other supported models) on behalf of customers. Billing is token-based.

4.2 Service Terms

• Billing: Per-token pricing; rates vary by model and usage tier

• Supported Models: OpenAI-compatible models, Mistral, DeepSeek, and others as listed on cloudidr.com

• Integration: API endpoint configuration

• Model Updates: cloudidr maintains model versions; customers notified of updates

• Data Handling: Customer prompts and responses are processed but not stored beyond operational logs (retained for 7 days)

• API Keys: Customer maintains their own API keys for model providers when applicable

4.3 Hosted Models Service Level Agreement

• Free Tier: No SLA commitment; best-effort service

• Paid Tiers: 99% API availability during billing period

• Response Time: 95th percentile response time within documented limits per model type

• Service Credits:

  • Below 99% availability: 10% credit

  • Below 95% availability: 25% credit

  • Below 90% availability: 50% credit

4.4 Customer Responsibilities

• Maintain valid API credentials

• Comply with model provider terms of service

• Monitor token usage and associated costs

• Ensure proper error handling in applications

5. LLM Ops

5.1 Service Description

LLM Ops is a cost tracking and monitoring service that provides visibility into AI API spending across multiple providers (OpenAI, Anthropic Claude, Google Gemini). The service operates as a pass-through proxy, logging metadata while routing API requests to the original providers.

5.2 How LLM Ops Works

• Proxy Service: LLM Ops sits between customer's application and AI providers (OpenAI, Anthropic, Google)

• API Key Handling: Customer API keys pass through the proxy via HTTPS but are never stored or logged

• Metadata Collection: Only metadata is collected (token counts, model names, timestamps, request IDs, calculated costs)

• Content Privacy: Customer prompts, responses, and all message content pass through encrypted but are never logged or stored

• Direct Routing: All API requests are immediately forwarded to the respective provider; cloudidr does not process or modify content

• Integration Options:

  • Direct proxy integration (change base URL)

  • Callback/webhook support for applications using LiteLLM or similar proxy frameworks

  • Custom integration via cloudidr API

5.3 LLM Ops Service Terms

6.3.1 Data Collection and Privacy

• Metadata Only: Cloudidr collects and stores only:

  • Token counts (input/output)

  • Model identifiers

  • Timestamps

  • Request/response IDs

  • Calculated costs

  • Custom tags (team, agent, department - if provided by customer)

• No Content Storage: Prompts, responses, and message content are never logged or stored

• API Key Security: API keys pass through via HTTPS but are never stored, logged, or retained

• Data Storage Location: All metadata stored in AWS data centers in the United States

• Data Retention: Metadata retained while customer account is active

• Data Deletion: All metadata permanently deleted within 30 days of account deletion

6.3.2 Service Availability and Reliability

• Best Effort: LLM Ops provided on best-effort basis for Free Tier

• Proxy Availability: If LLM Ops proxy becomes unavailable:

  • Customer may experience service interruption

  • Customer is responsible for implementing failover to direct provider access

  • cloudidr is not liable for unavailability or service interruptions

  • Customer can immediately route requests directly to providers by removing proxy configuration

• Data Accuracy: Cost tracking data provided on best-effort basis:

  • Cloudidr makes reasonable efforts to ensure accuracy

  • No guarantee of 100% accuracy; discrepancies may occur on cost calculations and other calculated metrics

  • Customers should verify costs against provider invoices

  • Cloudidr not liable for cost tracking errors or discrepancies

6.3.3 Pricing Tiers

• Free Tier (Current):

  • Unlimited API calls tracked

  • All supported providers (OpenAI, Claude, Gemini)

  • Real-time dashboard

  • Cost alerts

  • Team/agent/model breakdown

  • Limited features

  • No SLA commitment

  • Commercial use allowed

• Paid Tiers (Future):

  • Enhanced features (to be announced)

  • 99.99% uptime SLA

  • Priority support

  • Advanced analytics and optimizations

  • Custom integrations

  • Pricing to be announced

6.3.4 Free Tier Limitations

• No guaranteed uptime or availability

• No service credits for downtime

• Best-effort customer support

• Feature limitations compared to paid tiers

• cloudidr reserves right to add usage limits with 30 days notice

5.4 LLM Ops Service Level Agreement

6.4.1 Free Tier SLA

• No SLA Commitment: Free tier provided as-is with no uptime guarantees

• Best Effort: CloudIDR will make reasonable efforts to maintain service availability

• No Service Credits: Service credits not available for free tier

• Maintenance Windows: CloudIDR may perform maintenance without prior notice

6.4.2 Paid Tier SLA (When Available)

• Uptime Target: 99.99% monthly uptime

• Maximum Downtime: Approximately 4.38 minutes per month

• Measurement Period: Calendar month

• Exclusions:

  • Scheduled maintenance (with 48-hour notice)

  • Force majeure events

  • Customer's network or configuration issues

  • Third-party provider (OpenAI, Anthropic, Google) outages

  • DDoS attacks or security incidents

• Service Credits:

  • 99.9% - 99.99%: 10% credit

  • 99.0% - 99.9%: 25% credit

  • Below 99.0%: 50% credit

  • Below 95.0%: 100% credit

6.4.3 Data Accuracy

• No SLA or guarantee for cost tracking accuracy

• Best-effort calculation based on provider pricing

• Customer responsible for verifying costs against provider invoices

• Cloudidr not liable for cost calculation errors

5.5 Customer Responsibilities

• Integration: Properly integrate LLM Ops into applications (typically 2 lines of code)

• API Key Security: Maintain secure storage and handling of API keys

• Fallback Configuration: Implement fallback to direct provider access in case of proxy unavailability

• Cost Verification: Verify cost tracking data against actual provider invoices

• Usage Monitoring: Monitor API usage and implement appropriate rate limiting

• Compliance: Ensure use complies with provider terms of service (OpenAI, Anthropic, Google)

• Commercial Use: Allowed for all tiers; customer responsible for compliance with provider terms

5.6 Limitations of Liability (LLM Ops Specific)

In addition to Section 9 of the Customer Agreement:

• No Liability for Proxy Downtime: Cloudidr not liable for costs, losses, or damages resulting from LLM Ops proxy unavailability

• No Liability for Cost Tracking Errors: Cloudidr not liable for financial impact of cost tracking inaccuracies

• Third-Party Provider Responsibility: Cloudidr not liable for third-party provider (OpenAI, Anthropic, Google) issues, outages, or changes to their services

• Customer Failover Responsibility: Customer responsible for implementing failover mechanisms

• Data Loss: CloudIDR not liable for loss of metadata in case of service failure

7. Definitions

7.1 Instance-Minutes

Total minutes of usage per specific instance type. Example: m7g.xlarge used for 1.5 hours = 90 instance-minutes

7.2 Instance-Minutes Price

Calculated as: (per-minute price) × (instance-minutes)
Example: $0.50 per minute × 90 minutes = $45

7.3 Disaster Recovery (DR) Events

Qualifying Events:

• Physical infrastructure damage from:

  • Criminal activities

  • Military actions

  • Natural hazards

• Accidental fire or building damage

• Extended power outages exceeding backup capacity

• Cyber incidents causing infrastructure disability

Non-Qualifying Events:

• Temporary shutdowns from operator error

• Reversible software upgrade issues

• Issues resolvable through power cycling

7.4 Metadata (LLM Ops Specific)

Non-content information collected by LLM Ops including but not limited to:

• Token counts (input and output)

• Model names and identifiers

• Timestamps

• Request and response IDs

• Calculated costs based on provider pricing

• Custom tags provided by customer (team, agent, department)

Does NOT include:

• Prompts or queries sent to AI providers

• Responses or completions from AI providers

• Any message content

• API keys or authentication credentials

8. Service Credit Claims

To request service credits for SLA violations:

• Email: claims@cloudidr.com

• Required Information:

  • Account ID

  • Policy/Service name

  • Date and time of service failure

  • Description of impact

  • Supporting documentation

• Timeframe: Claims must be submitted within 30 days of incident

• Processing: CloudIDR will respond within 15 business days

9. Modifications to Terms

9.1 Right to Modify

• Cloudidr reserves the right to modify these Service Terms, including pricing, at any time.

• Changes may be made to reflect:

  • Service improvements

  • Market conditions

  • Legal or regulatory requirements

  • Operational needs

9.2 Notice of Modifications

• Material changes: 30 days advance notice

• Non-material changes: Effective upon posting

• Notice provided through:

  • Email to account administrators

  • Dashboard notifications

  • Service documentation updates

• Each notification will include:

  • Summary of changes

  • Effective date

  • Any required customer actions

9.3 Acceptance of Changes

• Continued use of Services after effective date constitutes acceptance of modified terms

• For material changes affecting services with commitments:

  • Existing policies/contracts continue under original terms until renewal

  • New or renewed services follow modified terms

• For pay-as-you-go services (FlexCompute, LLM Ops, GPU Service, Hosted Models):

  • Changes apply to all usage after effective date

  • No grandfather clauses for pay-as-you-go services

9.4 Customer Rights

• If you do not agree with material changes:

  • Complete current committed services under existing terms

  • Option to not renew under new terms

  • Cancel pay-as-you-go services without penalty

• Early termination options available per standard cancellation policies

9.5 Emergency Changes

• Immediate changes may be implemented if required for:

  • Security reasons

  • Legal compliance

  • Critical system stability

• Notice will be provided as soon as practicable

Document Version: 2.0
Previous Version Date: December 19, 2024

 Current Version Date: January 7, 2026

Summary of Changes from Previous Version

Major Additions:

1. Merged FlexCompute into Section 2: Schedule Compute Service (SCS) / FlexCompute with two modes (Scheduled Reservations and Pay-As-You-Go)

2. Added Section 3: GPU Service terms and SLA

3. Added Section 4: Hosted Models terms and SLA

4. Added Section 5: LLM Ops comprehensive terms, privacy, and SLA

5. Updated Section 7.4: Added Metadata definition for LLM Ops

6. Enhanced Section 9.3: Clarified acceptance rules for different service types

7. Added DRCS instance substitution policy in Section 1.2

8. Added LLM Ops webhook/callback integration support in Section 5.2

Service-Specific Changes:

• LLM Ops: Complete new section covering proxy operation, data privacy, no-storage policy, best-effort SLA, webhook/callback support

• SCS / FlexCompute: Merged into unified section with two modes (Scheduled Reservations + Pay-As-You-Go), AWS account integration

• GPU Service: New section covering SSH-only access model

• Hosted Models: New section covering token-based billing (removed time-based integration claims)

• DRCS: Added instance substitution policy for equivalent or higher instances

• All sections updated to reflect current service portfolio

Policy Changes:

• Clarified that API keys are never stored (LLM Ops)

• Established metadata-only collection policy (LLM Ops)

• Added best-effort accuracy disclaimer for cost tracking (LLM Ops)

• Added no-liability clauses for proxy downtime (LLM Ops)

• Clarified free tier has no SLA commitment (multiple services)

• Added future paid tier SLA commitments (LLM Ops, FlexCompute)