Pay for what you trace, not for seats
Every plan includes tracing, evals and guardrails. Invite your whole team. You pay for spans stored and evaluations run.
Hobby
For side projects and your first agent in production.
Free forever
- 50,000 spans / month
- 1,000 eval runs / month
- 14-day retention
- 3 guardrail rules
- Community support
Team
For product teams running several agents with real users.
Includes 2M spans, then $0.80 per 100k
- Everything in Hobby, plus
- 2M spans / month included
- 50,000 eval runs / month
- 90-day retention
- Replay and CI gates
- Unlimited seats and guardrails
- Slack alerts, shared Slack channel
Enterprise
For regulated industries and agents at very high volume.
Annual contract, volume pricing
- Everything in Team, plus
- Self-hosted or single-tenant cloud
- Up to 400-day retention
- SSO, SCIM and audit logs
- EU and AU data residency
- 99.95% uptime SLA, named engineer
Estimate your bill
A typical support agent produces 8–15 spans per conversation. Drag to your monthly volume.
Above 50M spans a month, Enterprise volume pricing is usually cheaper. Ask for a quote →
Compare plans
All plans include every product. Limits and support are what change.
| Hobby | Team | Enterprise | |
|---|---|---|---|
| Tracing | |||
| Included spans | 50k | 2M | Custom |
| Retention | 14 days | 90 days | Up to 400 days |
| Sessions & user tracking | ✓ | ✓ | ✓ |
| PII redaction in SDK | ✓ | ✓ | ✓ |
| Evals | |||
| Eval runs / month | 1,000 | 50,000 | Custom |
| Judge calibration | — | ✓ | ✓ |
| Replay | — | ✓ | ✓ |
| CI gates | — | ✓ | ✓ |
| Security & support | |||
| SSO / SCIM | — | Google only | ✓ |
| Self-hosting | — | — | ✓ |
| Support | Community | Shared Slack | Named engineer, 1h SLA |
Questions
Something else? Email billing@sightline.dev
What counts as a span?
One step in a trace: a model call, a tool call, a retrieval or a custom block you mark with @sightline.step. Child HTTP requests inside a tool call are free.
What happens if I go over my plan?
On Hobby, new spans are sampled at 10% until the month resets. We never drop errors. On Team, overage is billed at $0.80 per 100,000 spans and you can set a hard cap in settings.
Do LLM-judge evals use my model credits?
By default, judges run on our models and are included in your eval runs. You can bring your own API key to use a specific model. Those calls are billed by your provider.
Can I self-host?
Yes, on Enterprise. We ship a Helm chart and Terraform modules for AWS, GCP and Azure. Most teams are running within a week.
Do you offer startup or nonprofit discounts?
Companies under 2 years old with less than $5M raised get Team free for 12 months. Nonprofits get 50% off every plan.