01
Dev
Try it on a real agent
Free 25k spans a month included
Start free - Every step recorded
- Scored and risk-checked live
- Hold, approve or block
- SDK install in minutes
Pricing · Usage-based Observability, evaluation and enforcement.
Observability, evaluation and enforcement.
In real time.
Every plan is the full platform. Pay for what your agents do.
Plans
save 20% by switching to annual
All prices USD
02 Most popular
Scaleup
Agents in production
From $9,600 USD / year, 400k spans a month included commit more at $2.00 per 1k, up to 4M a month
From $250 USD / month, first 100k spans included then $2.50 per 1k up to 4M a month, cancel any time
Get started - Everything in Dev
- 100% of activity, real time
- Dev, staging and prod
- Unlimited seats
03
Enterprise
4M+ spans a month
Custom rates step down with scale
Talk to us - Everything in Scaleup
- SSO and audit retention
- SLA and named engineer
- Pay by invoice or PO
Every plan includes
- Every call, tool, and decision recorded and risk-scored live
- Real-time enforcement to hold, approve, or block
- PII checks built into the reliability loop
- Dev, staging, and production with eval gates
- Agent versioning with instant rollback
- Unlimited seats on every plan
Estimate
What would your agents cost?
One span is one step your agent takes.
- An LLM call: one span
- A tool call: one span
- A message turn: one span
- A custom business step you mark: one span
Scores, risk checks, and interventions never create spans.
spans / month1,000,000
monthly billing$2,500 USD / mo
annual, committed$24,000 USD / yr
save $6,000 / yr on annual
enterprisetalk to us, rates step down
| Dev | Scaleup | Enterprise | |
|---|---|---|---|
| Observability | |||
| Every step recorded: each call, tool and decision | • | • | • |
| Nested and custom business spans | • | • | • |
| Agent versioning and rollback | • | • | • |
| Dev, staging and prod environments | • | • | • |
| Evaluation | |||
| Real-time scoring on 100% of activity | • | • | • |
| Risk, pattern and sequence checks | • | • | • |
| PII checks, no per-check cost | • | • | • |
| Latency P95 / P99 and cost per step | • | • | • |
| Optional BYO LLM evals | • | • | • |
| Enforcement | |||
| Hold, approve or block in real time | • | • | • |
| Eval-gated promotion | • | • | • |
| Team and deployment | |||
| Unlimited seats | • | • | • |
| Self-hosted deployment | • | ||
| SSO and extended audit retention | • | ||
| Pay by invoice or PO, SLA, named engineer | • | ||
The meter
How the meter works
One step, one span
A span is one step your agent takes: an LLM call, a tool invocation, a message turn, or a custom business step you mark. The SDK records them automatically as your agent runs.
Every span scored and enforced
Deterministic scoring, risk checks, and runtime enforcement run on 100% of activity with no per-check cost. Deterministic scoring uses zero tokens, and scores and checks never create spans.
No token markup
Your models, keys, and rates stay yours. Prefactor never resells or marks up model usage. Optional LLM evaluations use your own model account.
No seat math
Seats are unlimited on every plan. Add every engineer, reviewer, and stakeholder without changing the invoice.
Security
Every plan runs with the same protections
Encrypted end to end
AES-256 encryption at rest and TLS 1.3 in transit, for all data and audit logs, on every plan.
Keys scoped per environment
Your account and each environment (dev, staging, and prod) has its own API keys, so access stays isolated end to end.
PII detection as standard
PII checks run on every span, on every plan, inside the span price: detected and redacted, blocked, or escalated in real time.
Your data never trains models
Prefactor never uses agent interactions or customer data to train, fine-tune, or improve any model. Your data stays yours.
Immutable audit trail
Every agent action, score, and intervention is logged to append-only storage: a tamper-evident record you can hand to an auditor as proof of what your agents did.
Audit-ready
SOC 2 Type II is in progress, and risk mapping is built for GDPR. HIPAA, ISO 27001, NIST AI RMF, and EU AI Act controls are built in, with data residency in the US, EU, and Asia Pacific.
Full detail on infrastructure, runtime controls, and responsible disclosure: /security →
Questions
Pricing questions
What counts as a span?
One step your agent takes: an LLM call, a tool call, a message turn, or a custom business step you mark. The SDK records them automatically as your agent runs. Scores, risk checks and interventions attach to spans and never create new ones, so the checking is free.
Monthly vs annual, what changes?
Rate and commitment only. Monthly is $250 a month, which includes your first 100,000 spans, then $2.50 per 1k, cancel any time. Annual commits a monthly volume at $2.00 per 1k from 400k spans a month. Learn your volume on monthly, then commit it.
What if I exceed my annual commitment?
Nothing breaks. Overage bills at $2.50 per 1k and you can raise the commitment at renewal. Self-serve covers up to 4M spans a month.
I do not know my volume yet. Where do I start?
Monthly. Worst case is $250 and your first month tells you your real number.
Do checks and interventions cost extra?
No. Scoring, risk analysis, PII checks and interventions run on every span, in real time, inside the span price. There is no per-check meter.
Do you charge per seat or mark up tokens?
Neither. Unlimited seats on every plan, and optional LLM evals are bring your own model at your rates.
What currency are prices in?
USD. Billing is handled by Stripe and your card is charged in US dollars.
Can I self-host?
Yes, on Enterprise plans, billed on committed volume.
Get started