Question 1

Can I trust DriftWatch with my provider API keys?

Accepted Answer

Your provider keys are encrypted at rest, scoped to a single environment, and never written to logs, analytics, or support tooling. We run zero-data logging on prompt and completion bodies by default — payloads are processed in memory and dropped when the request completes. DriftWatch-issued keys are separately scoped and revocable instantly, so you can rotate access without touching provider credentials. Full details, including our SOC 2 Type II and ISO 27001 compliant infrastructure posture, are on the Security page.

Question 2

How does proxy latency hold up at high volume?

Accepted Answer

Median proxy overhead is under 15ms globally (about 12ms at the edge region closest to your provider), and it stays flat under sustained load thanks to multi-region anycast routing and active-active failover. Because pruning reduces the number of input tokens the model has to read, most teams see a net improvement in time-to-first-token rather than a slowdown.

Question 3

What makes DriftWatch different from open-source logging tools?

Accepted Answer

Open-source loggers record what already happened: requests, tokens, and latency. DriftWatch acts on the request in flight — it prunes token bloat before the call leaves the edge (30-40% typical input reduction), validates every response against your JSON contract with retry and failover, and scores model behavior against a rolling baseline so silent provider-side drift becomes an alert instead of a production incident. Logging tells you the bill; DriftWatch lowers it and protects the contract.

Question 4

How do you compare to Helicone or Portkey?

Accepted Answer

Helicone focuses on open-source logging, analytics, and caching; Portkey focuses on multi-provider routing and key management. DriftWatch is a context firewall and drift monitor: active token pruning plus schema enforcement and drift scoring on every response. See our detailed side-by-side comparisons at /vs/helicone and /vs/portkey.

Question 5

What happens after I purchase the Enterprise plan?

Accepted Answer

Immediately after checkout, our control plane provisions a dedicated single-tenant edge proxy instance exclusive to your workspace. Within 60 seconds your custom endpoint is live at api.driftwatchproxy.com/v1/ent-[customer-id] and ready to accept traffic. No tickets, no onboarding calls, no manual DNS work.

Question 6

How long does integration take?

Accepted Answer

Most teams are routing production traffic in under 2 minutes. Change one line — your SDK baseURL — and keep your existing API keys, request shapes, and response handling. We provide copy-paste snippets for OpenAI (Python/TypeScript), Anthropic (Python/TypeScript), and raw cURL.

Question 7

Can I run DriftWatch in a single region or multi-region?

Accepted Answer

Enterprise deployments include automatic multi-region anycast routing by default. Traffic is served from the edge location closest to your provider, with optional region pinning for data-residency requirements.

Question 8

What uptime SLA does Enterprise include?

Accepted Answer

Enterprise Platform includes a 99.99% monthly uptime SLA backed by automated health checks and multi-region failover. If our availability drops below the SLA, we automatically credit your account per the published service-level terms.

Question 9

How much latency does the proxy add?

Accepted Answer

Median overhead is under 15ms globally, with 12ms at the edge region closest to your provider. Because prompt pruning reduces the number of tokens the model must read, we often observe net time-to-first-token (TTFT) speedups rather than slowdowns.

Question 10

What happens if the proxy fails?

Accepted Answer

Every Enterprise instance is deployed with active-active failover across cloud regions. If an edge location degrades, traffic is rerouted automatically. You can also configure a backup provider endpoint (for example, Claude 3.5 Sonnet) in our dashboard for automatic drift-driven failover.

Question 11

Do you store or train on our prompts?

Accepted Answer

Never. We do not train on customer traffic. Our zero-retention mode processes prompt bodies in memory only and drops them the moment the request completes. Metadata logs are optional and can be disabled per environment. All payload routing is end-to-end encrypted.

Question 12

How do SOC2 and HIPAA audit exports work?

Accepted Answer

Enterprise users can export filtered telemetry logs in one click as CSV or JSON. The export includes request metadata, schema validation outcomes, latency measurements, and token counts — everything your compliance team needs, with no manual data collection.

Question 13

Is the proxy suitable for regulated workloads?

Accepted Answer

Yes. Enterprise single-tenant instances keep your traffic isolated, and our audit-log exporter is designed for SOC2, HIPAA, and ISO 27001 evidence collection. Custom BAA and DPA workflows are self-serve from the Security page.

Question 14

How are overages calculated?

Accepted Answer

Each tier includes a base token allocation (750M tokens / month on Enterprise Platform, at $9,999 / month or $101,988 / year billed annually). Usage beyond that is billed per 1,000 tokens at the plan's overage rate: $0.020 on Pro, $0.012 on Growth, and $0.006 on Enterprise. Overage charges are itemized on your invoice and updated daily in the dashboard.

Question 15

Can I set a hard monthly spend cap?

Accepted Answer

Yes. Every plan includes self-service spend guardrails. You can set a hard cap per environment; when the cap is reached, traffic passes through unpruned rather than incurring unplanned charges. Enterprise plans also support custom overage thresholds and team-level budgets.

Question 16

Can I switch between monthly and annual billing?

Accepted Answer

Absolutely. Upgrades take effect immediately and are prorated to the day. Annual billing saves roughly 17% compared to monthly. Downgrades apply at the start of your next billing cycle so you keep the retention window you already paid for.

Feature	Developer	Pro	Growth	Enterprise
Monthly tokens included	250,000 (hard cap)	15M	150M	750M
Overage rate / 1,000 tokens	Not included	$0.020	$0.012	$0.006
API keys	1	5	25	Unlimited
Team seats	1	5	15	Unlimited
Token pruning engine	Not included	In-line AST pruning	In-line AST pruning	In-line + custom profiles
Schema firewall rules	3 basic rules	25 rules, real-time	Unlimited + auto-failover	Unlimited + health diagnostics
Model drift detection	Alerts only	Alerts + baselines	Baselines + auto-failover	Baselines + custom thresholds
Log retention	7 days	30 days	90 days	1 year (configurable)
Support SLA	Community FAQ	24-hour email	4-hour priority	1-hour priority
Uptime SLA	Best effort	99.9%	99.95%	99.99% with credits
Single-tenant dedicated instance	Not included	Not included	Not included	Included, auto-provisioned
SOC 2 / HIPAA audit log export	Not included	Not included	CSV export	One-click CSV + JSON
Spend guardrails & hard caps	Hard cap only	Yes	Yes, per environment	Yes + team budgets

Simple, predictable infrastructure pricing.

Token Savings Calculator

Why we cost more than basic observability tools

Developer

Pro (Startup)

Growth (Mid-Market)

Enterprise Platform

Compare every plan

Enterprise & Security FAQ

Ready to scale your LLM infrastructure?