Ready-to-use SLA templates for different deployment scenarios. Customize these for your specific needs.
01Section
How to Use These Templates
- Choose the template matching your scenario
- Fill in bracketed fields
[like this] - Review and adjust targets based on your baselines
- Have legal review before external use
- Set up monitoring to track compliance
02Section
Template 1: Basic Agent SLA
For internal tools or early-stage products
snippet
SERVICE LEVEL AGREEMENT
[Agent Name]
1. AVAILABILITY
Target: 99%
Measurement: HTTP 2xx responses every 15 minutes
Period: Calendar month
2. PERFORMANCE
TTFB P50: <5 seconds
3. EXCLUSIONS
- Scheduled maintenance with 24 hours notice
- Customer-caused issues
- External dependencies outside Provider's control
4. REMEDIES
This SLA is informational. No service credits apply.03Section
Template 2: Standard Agent SLA
For production customer-facing agents
Service Level Objectives
| Metric | Target |
|---|---|
| Availability | 99.5% |
| Eval Pass Rate | 95% |
| TTFB P50 | <3 seconds |
| TTFB P95 | <6 seconds |
| Per-Region Availability | >95% |
Service Credits
| Actual Availability | Credit (% of Monthly Fee) |
|---|---|
| 99.0% - 99.5% | 10% |
| 98.0% - 99.0% | 25% |
| 95.0% - 98.0% | 50% |
| Below 95.0% | 100% |
Correctness credits: 10% per 1% below Eval Pass Rate target. Credit cap: 50% of monthly fee.
Key Provisions
- Monitoring by third-party service (Agent Status or equivalent)
- Check frequency: every 15 minutes from 3+ regions
- Monthly compliance reports by 5th business day
- Termination right if target missed for 3 consecutive months
- 48 hours maintenance notice, max 4 hours/month
04Section
Template 3: Premium Agent SLA
For mission-critical enterprise deployments
Service Level Objectives
| Metric | Target |
|---|---|
| Availability | 99.9% |
| Health Eval Pass Rate | 99% |
| Contract Eval Pass Rate | 98% |
| TTFB P50 | <2 seconds |
| TTFB P95 | <4 seconds |
| TTFB P99 | <8 seconds |
| Primary Region Availability | 99.5% |
| Secondary Region Availability | 98% |
Incident Response Requirements
| Severity | Acknowledge | Update Frequency | Target Resolution |
|---|---|---|---|
| S1 (Critical) | 15 min | 30 min | 2 hours |
| S2 (High) | 30 min | 1 hour | 4 hours |
| S3 (Medium) | 2 hours | 4 hours | 24 hours |
| S4 (Low) | 24 hours | Daily | 5 business days |
Enhanced Credits
- Availability: 10-20% per 0.1% below target (tiered)
- Correctness: 5% per 0.5% below eval pass rate target
- Performance: 10% per month TTFB P95 exceeds target
- Incident response: 5% per incident exceeding response time
- Per-month cap: 100% of monthly fee
- Annual cap: 300% of annual fee
Termination Rights
Customer may terminate immediately if two S1 incidents occur in any 30-day period, availability falls below 98% for any month, or material breach is not cured within 30 days.
05Section
Template 4: API Product SLA
For agents offered as an API product to developers
Plan Commitments
| Plan | Availability | TTFB P50 | TTFB P95 | Rate Limit |
|---|---|---|---|---|
| Starter | 99.0% | <5s | <10s | 10 req/s |
| Professional | 99.5% | <3s | <6s | 50 req/s |
| Enterprise | 99.9% | <2s | <4s | Custom |
Credits
| Availability | Credit |
|---|---|
| Target - 1% | 10% |
| Target - 2% | 25% |
| Target - 5% | 50% |
| Below Target - 5% | 100% |
Support Response Times
| Plan | Support Channels | Response Time |
|---|---|---|
| Starter | 48 hours | |
| Professional | Email, Chat | 24 hours |
| Enterprise | Email, Chat, Phone | 4 hours |
Exclusions
- Free tier usage
- Scheduled maintenance (status page notifications)
- Customer-side issues (invalid requests, auth failures)
- Rate limiting within plan limits
- Beta or preview features
- Force majeure
06Section
Usage Notes
Before Using
- Review targets against your actual capabilities
- Measure baselines before committing
- Have legal counsel review external-facing SLAs
- Set up monitoring to track compliance
Customization Points
- Adjust percentage targets based on actual performance
- Modify credit structures based on business model
- Add or remove geographic regions as relevant
- Customize incident severity definitions
