Requesty Pricing: Free Tier, 5% Markup, and Enterprise Scope

Requesty pricing publishes a clear commercial formula. The Free tier covers free models within a daily request allowance. Pay as you go adds 5% to model cost with no stated minimum spend or per-seat charge. Enterprise keeps the gateway model and adds organization controls with support. Custom service commitments are also available. The budget should preserve provider consumption and Requesty's margin as separate lines.

TL;DR

• Requesty lists a Free tier with access to free models and 200 requests per day.
• Pay as you go charges the underlying model cost plus a 5% markup, with no published minimum spend or per-seat fee.
• Enterprise adds SSO and role-based access. Dedicated support and custom SLAs are available on request.
• Routing and caching appear across the published plans. EU data residency does too, but buyers should verify the current order and measurement rules.

Requesty pricing uses three commercial paths

The Requesty pricing page presents Free and Pay as you go alongside Enterprise. Free lists access to free models and 200 requests per day. It requires no credit card. Pay as you go lists a 5% markup on model cost and says there is no minimum spend. Enterprise directs buyers toward a custom conversation and lists dedicated support plus custom SLAs.

The page also gives a concrete illustration: a model priced by its provider at $10 per 1 million tokens would cost $10.50 through Requesty under the 5% formula. That example explains the arithmetic, but the underlying provider rate can change by model and input type. Cache discounts or other model-specific pricing can change it too.

I prefer this formula to a vague "contact sales" page. Finance can see the gateway margin. Engineering still has to explain which model calls created the base cost.

Provider cost and markup need separate ledgers

A monthly Requesty total combines two concepts: the provider model has a usage price, while Requesty applies its published margin for the pay-as-you-go route. Record the provider and model beside the usage category, effective provider rate, and 5% gateway charge.

That separation matters when routing changes. Moving one workload to another provider can alter the base model cost while the Requesty percentage remains the same. A fallback can select a different destination during an outage. Cached responses can also change how often an upstream model call occurs.

The model routing cost guide explains why a route decision needs operational evidence beside the commercial schedule. Put the Requesty usage export on one monitor and the provider model catalog on another. The useful row identifies the application and route. It also records the selected model, token usage, cache state, and final charge.

Free and pay-as-you-go tiers fit different tests

The published Free tier is a useful development entry point because it covers free models and sets a daily request allowance. It should be treated as an entitlement with a boundary, not as a forecast for a production workload using paid models.

Pay as you go fits teams that want a variable bill tied to model consumption. The page says routing and caching are included on every plan. It also lists EU data residency. Buyers should verify the current feature matrix and ask how spend limits and analytics behave. They should also ask about guardrails and MCP Gateway access on the selected tier.

The Requesty gateway page describes one OpenAI-compatible endpoint with model routing and response caching. It also describes automatic failover and observability. Those functions can change realized cost. A cache hit may avoid an upstream call. A failover may choose a provider with a different rate. Preserve the actual route rather than budgeting from the configured default alone.

Enterprise scope adds identity and governance

The Requesty Enterprise page describes SSO and role-based access across organization, group, service-account, user, and key layers. It also describes approved-model controls. It also names an EU gateway in Frankfurt and a DPA on request. Dedicated support and custom SLAs appear on the pricing page.

These controls create quote inputs beyond token consumption. Count human users and service accounts. List groups and applications, then identify each key. Define approved model sets and spend limits. State the required data region and retention period. Specify the audit export and zero-data-retention conditions. Add support coverage and a written availability commitment when the gateway sits on a production path.

An illustrative dashboard on a product page is not a customer quote. The dollar values shown inside interface mockups should never be copied into a budget as published package prices. Use the written pricing formula and the signed Enterprise order.

Reconstruct the request behind the bill

Requesty's routing and observability records support cost operations. An authorization review may need more context about the originating user or workload and the approved destination. It may also require the data classification, policy version, and outcome. Those fields should be tested on the selected plan instead of inferred from words such as governance or audit.

The AI gateway and LLM router comparison separates destination selection from enterprise authorization. Requesty can centralize model access and route traffic. It can also cache responses and fail over. A buyer should still run one denied-request test and inspect the exported record. Change one attribute, such as the approved model list, then verify the decision and identity that appear in evidence.

Put each cost on its own line. Requesty's 5% margin belongs with routing and gateway services. Provider model usage remains the base consumption. Any independent policy layer and SIEM storage deserve separate entries. Review labor does too.

DeepInspect

DeepInspect can sit before a Requesty HTTP endpoint when authenticated user or agent traffic is deliberately routed through its proxy. It evaluates application-supplied identity and role against the classification and destination, then applies versioned policy and records the decision before forwarding allowed traffic.

Requesty remains responsible for its markup and model catalog. It also remains responsible for routing, caching, and failover. DeepInspect adds an identity-bound authorization record at the model request boundary without changing Requesty's published price formula. Book a technical deep dive at deepinspect.ai.