R

LLM gateways

Requesty AI Gateway Review

requesty.ai · checked 21 Oct 2026

Managed AI gateway with a shared OpenAI-style API for 600+ models, routing, analytics and governance.

How it plugs in

—Public API
NoOpen source
YesFree tier

Signal on LLM gateways6.8#10 of 33

Requesty puts a common, OpenAI-compatible interface in front of over 600 models supplied by more than 30 providers. It adds routing, fallbacks, caching, usage analytics, budgets and governance controls. Requesty runs through web, API and extension access, with free, usage-based and custom enterprise plans.

Requesty home page

Review

In a line

Requesty provides a hosted way to reach over 600 language models through a single API that follows OpenAI’s interface.

It runs as a hosted service rather than a self-hosted server.

What it does

The service places one request format in front of models from more than 30 providers. Existing OpenAI Python and JavaScript clients can point to it by replacing the endpoint and key, so a separate Requesty SDK is unnecessary.

Policies can direct traffic by geography, response speed or assigned weights. They also support retries on failed or timed-out calls, backup-model failover, weighted distribution and nesting. Automatic prompt caching is available too, and Requesty claims repeated context may cost up to 90% less.

Requesty tracks model, team, user and key costs, latency and usage. Spending limits and notifications can be set for keys, individual users, teams or whole organizations. Versioned prompts can be shared and changed without a deployment.

How it connects

Requesty names OpenAI, LangChain, PydanticAI, Vercel AI SDK, Haystack and LlamaIndex TS among its SDK and framework connections. The integrations page names additional tools such as Claude Code, Cline, Roo Code, VS Code, GitHub Copilot, Codex, n8n, LibreChat and OpenWebUI.

Its interface includes chat, embeddings, audio, image and model-listing endpoints, while management endpoints handle keys, groups, members and organization settings. Requesty also publishes a Python library, Vercel AI SDK package, CLI, LibreChat integration and n8n node; the CLI is MIT-licensed.

For broader context, see LLM gateways. Requesty offers regional endpoints, including an EU gateway in Frankfurt, and EU requests can be routed there.

Good fit

Requesty suits teams that want to switch among many language models while keeping one application endpoint. Its routing, fallback and load-balancing policies can reduce the need to build those controls into each app.

It is also suited to organizations that need shared budgets, model allowlists, team roles, audit logs, PII detection, content guardrails and cost reporting.

Watch for

Requesty is managed only; self-hosting is not offered. On self-serve plans, encrypted EU storage retains prompt and output logs for as long as 30 days by default. Logging can be disabled for a particular key, and the organization can request zero retention in writing.

The site says its SOC 2 Type II program is in progress. A GDPR Article 28 data-processing agreement is available on request. Check your retention, residency and compliance needs before sending sensitive prompts.

Pay as you go has a 5% markup on model cost, and its listed price is not provided. Enterprise plans use custom quotes.

Pricing

The no-cost Free plan requires no card and offers free models, 200 daily requests, routing, caching, fallbacks, spend analytics and EU residency.

Pay as you go bills according to usage, with model charges increased by 5%.

Enterprise is available through a custom quote. It adds SSO/OIDC, role-based access and audit logs, approved models, teams and budgets, guardrails, PII detection, service accounts, dedicated support and custom service-level agreements. Requesty also offers a free trial.

Verdict

Choose Requesty if your software needs a common interface for many AI providers and you value policy-based routing, automatic backups, spend limits and centralized visibility. It can also fit teams that need EU routing and governance controls without running their own gateway.

Avoid it if you require a self-hosted deployment, a published Pay as you go price or a completed SOC 2 Type II program. Review default logging and the 5% usage markup before committing, especially for sensitive or high-volume workloads.

Our take

Requesty is a practical choice for teams standardizing access to many AI models through one endpoint. The free plan lets you try routing, caching and analytics without a card, while paid options add broader model access and enterprise controls. Its hosted-only setup simplifies operations but removes self-hosting. Buyers handling sensitive data should review default log retention, EU routing and the in-progress SOC 2 program. Budget for the 5% Pay as you go markup, or request an Enterprise quote if you need custom support and controls.

  • A single OpenAI-style endpoint reaches more than 600 models
  • Routing, failover, load distribution and prompt caching
  • Cost, latency and usage analytics with budget alerts
  • PII detection, guardrails, roles, allowlists and audit logs
  • No-cost plan with 200 daily requests and no card
  • No self-hosted deployment
  • Usage-based Pay as you go billing carries a 5% model-cost markup
  • Self-serve prompt and output logging starts enabled by default
  • SOC 2 Type II program is still in progress
  • Pick it ifTeams comparing or routing requests across many AI models; Organizations needing shared budgets and governance controls; Apps already using OpenAI-compatible SDKs
  • Skip it ifYou need to operate the gateway on your own infrastructure; You need a fixed published price for usage-based billing; Completed SOC 2 Type II is mandatory for your compliance policy

Pricing

The Free option includes 200 requests daily without a credit card. Pay as you go follows usage and applies a 5% markup to model cost; no plan price is listed. Enterprise plans are quoted individually. Requesty also offers a free trial.

  • Freeno credit card required · free models · 200 requests/day · routing, caching and fallbacks · spend analytics · EU residencyFree
  • Enterprisecustom pricing · SSO/OIDC · RBAC and audit logs · approved models · teams and budgets · guardrails and PII detection · service accounts · dedicated support · custom SLAsOn request
  • Pay as you go5% markup on model cost; usage based · 600+ models · BYOK · routing policies · caching and fallbacks · spend caps · MCP Gateway · EU residency · advanced observability · email supportOn request

As requesty.ai lists them, checked 21 Oct 2026.

Questions

Is Requesty an AI model?

No. It is a hosted AI gateway that provides one endpoint for models from more than 30 providers.

Can existing OpenAI code work with Requesty?

Yes. Existing official Python or JavaScript OpenAI clients can use Requesty after you change the base URL and API key.

Does Requesty offer a free plan?

Yes. Requesty’s Free tier costs nothing, needs no card and allows 200 requests per day.

Can Requesty be self-hosted?

No. Requesty operates as a managed service and does not provide a self-hosting option.

Where are EU requests processed?

Requesty identifies Frankfurt as its EU gateway location and says EU requests can be routed there.

How long are prompts and outputs kept?

On self-serve plans, encrypted EU logs are kept for up to 30 days by default; a key-level setting can turn logging off.