Requesty
requesty.ai
- Free plan
- Web
- Browser extension
- API
What Requesty is
Requesty is a managed AI gateway that connects applications to more than 600 models through one API. The API is OpenAI-compatible, so existing SDK code can work after changing the base URL and API key. Routing includes policy and regional rules, automatic failover, and load balancing. Teams can monitor latency and usage, and track costs with analytics by model, user, and team. Governance tools include PII detection and scrubbing, content guardrails, team roles, model allowlists, and audit logs. Integrations include OpenAI, LangChain, PydanticAI, Vercel AI SDK, Haystack, and LlamaIndex TS, alongside tools such as Claude Code, Cline, VS Code, and n8n. Requesty offers web and API access but not self-hosting. The Free plan costs nothing and includes 200 requests per day. Pay as you go costs $5 a month, billed with a 5% markup on model cost. Enterprise pricing is custom. Self-serve prompt and output logging is enabled by default and retained encrypted in the EU for up to 30 days; it can be disabled per key.
Who it suits
Requesty suits developers and teams that want to route model requests through one API and monitor usage, costs, and governance. It is not suited to organizations that require self-hosted deployment.
Good to know, and what to watch
- Access to 600+ models through one API.
- OpenAI-compatible API supports existing SDK code.
- Routing includes failover and load balancing.
- Analytics track costs by model, user, and team.
- Free plan includes 200 requests per day.
- Self-hosted deployment is not offered.
- Self-serve prompt and output logging is on by default.
- Pay as you go applies a 5% markup on model cost.
Plans
Every plan and priceEnterprise
Pay as you go
Free
Free plan
Getting started
- Open the Requesty website or API documentation
- Choose Free, Pay as you go, or Enterprise
- Create an API key
- Point compatible SDK code to Requesty’s base URL
- Configure routing, model access, and spend controls
- Connect a listed SDK, framework, or agent tool
Limits to check
Free permits 200 requests per day and includes access to free models. Pay as you go charges a 5% markup on model cost; Enterprise pricing is custom.
The Club's review
Requesty is a hosted AI gateway for developers and teams working across many language models. It is best for organizations that want routing, cost oversight and access controls without operating gateway infrastructure. The free tier is useful for small-scale evaluation, while production teams should weigh the usage markup and managed-only deployment.
Overview
Requesty provides one OpenAI-compatible API to more than 600 models from 20 providers. Teams using compatible SDKs can generally point existing code at Requesty by changing the base URL and API key, avoiding a rewrite for each model provider. API and web access cover both application integration and browser-based administration.
The service centralizes model access, traffic decisions and governance. That is a practical alternative to maintaining separate provider integrations, but it also means Requesty remains part of the request path and customers cannot run the gateway themselves.
Key features
Routing and reliability
Policy-based routing lets teams direct requests according to their own rules, while regional routing can send European traffic through Frankfurt. Automatic failover and load balancing help distribute requests and keep an application moving when a route becomes unavailable. Fallback routing and caching are included even on the Free plan, making the entry tier more capable than a simple model directory.
Observability and controls
Real-time analytics show cost by model, user and team, with latency and usage monitoring for operational review. Pay as you go adds spend limits and budget caps, which are more useful for teams managing shared production consumption than for an individual experimenting with prompts.
Governance features include PII detection and scrubbing, content guardrails, team roles, model allowlists and audit logs. Enterprise adds full RBAC, approved policies, groups and spend controls, making it the appropriate tier when access decisions need formal administration. Virtual API keys support separating users or workloads without handing out a single organization credential.
Integrations and security
Named integrations cover OpenAI, LangChain, PydanticAI, Vercel AI SDK, Haystack and LlamaIndex TS. The integration catalogue also includes Claude Code, Cline, Roo Code, VS Code, GitHub Copilot, Codex, n8n, LibreChat and OpenWebUI, which broadens its usefulness beyond a single application stack.
Traffic uses TLS 1.2 or higher and stored data uses AES-256. Prompt and output logging is enabled by default on self-serve plans and retained for up to 30 days in encrypted EU storage; logging can be disabled per key, while organization-wide zero data retention requires a written request. SOC 2 Type II is still in progress, and a GDPR Article 28 DPA is provided on request.
Pricing
Free
The Free plan costs nothing and requires no credit card. It includes access to all free models, 200 requests per day, routing, caching, fallbacks, spend tracking and analytics, plus EU data residency. The daily request ceiling makes it suitable for prototypes, evaluation and low-volume internal tools, not sustained production traffic.
Pay as you go
Pay as you go costs $5 a month, with a 5% markup on model cost and charges based on usage. It unlocks all 600-plus models, bring-your-own keys, routing policies, caching and fallbacks, spend limits, budget caps, the MCP Gateway, EU residency, advanced observability and email support. The fixed monthly fee plus markup is a meaningful cost over direct provider access, but the controls can justify it for teams that need centralized accounting and routing.
Enterprise
Enterprise uses custom pricing. It adds SSO, full RBAC and audit logs, approved models and policies, teams and groups, spend controls, guardrails, PII detection, EU residency, CI/CD service accounts, dedicated support and custom SLAs. It is the sensible choice for regulated or larger organizations that need formal identity, policy and support commitments.
Platforms
Requesty is offered through API and web interfaces. It is a managed service rather than a deployable gateway, so self-hosted operation is not available.
Who it's for
Choose Requesty if your team wants one compatible endpoint for many models and values routing, failover, analytics and centralized governance in a hosted platform. Developers can preserve much of an OpenAI-style integration, while platform teams gain cost and policy controls. Look elsewhere if self-hosting, local infrastructure ownership or avoiding a model-cost markup is a primary requirement.
Pros and cons
- Pros: One OpenAI-compatible API spans more than 600 models and 20 providers, reducing integration work.
- Pros: Policy and regional routing, failover, load balancing and caching address both reliability and operational control.
- Pros: Cost analytics by model, user and team make shared usage easier to manage.
- Pros: EU routing and encrypted EU retention support organizations with regional data requirements.
- Cons: Self-serve prompt and output logging is on by default, so teams must change key settings or request organization-wide zero retention.
- Cons: Pay as you go combines a $5 monthly fee with a 5% model-cost markup.
- Cons: The Free plan's 200 requests per day limits its production usefulness.
- Cons: Self-hosting is unavailable, and SOC 2 Type II remains in progress.
Alternatives
See LLM Gateway Software for a broader shortlist. Choose Bifrost instead when self-hosting through npx, Docker, Kubernetes or a Go binary and Apache 2.0 licensing matter. Kong Gateway suits teams wanting a conventional gateway with a 30-day trial and no gateway limits during that period. LLM Gateway is an option for a freemium service covering 200-plus models across 40-plus providers. TrueFoundry AI Gateway fits small teams seeking a free developer tier with user, request and model-account caps. Helicone is worth considering if an open-source gateway and a free request allowance are priorities. LiteLLM is the stronger choice for teams that require an open-source, self-hosted gateway. Vercel is a broader web platform rather than a direct managed model gateway. Envoy AI Gateway suits organizations wanting a free, self-managed project for Linux, macOS or Kubernetes.
Verdict
Requesty is a strong fit for teams that want broad model access, routing and governance through one managed API, with EU residency and useful analytics included from the lower tiers. Its main reasons to look elsewhere are equally clear: self-hosting is not offered, self-serve logging is enabled by default, and the paid entry plan adds both a subscription and a model-cost markup.
The verdict
Requesty brings model routing, monitoring, and governance into a managed gateway, with a free plan and a pay-as-you-go option. Teams should account for the default logging on self-serve plans and the lack of self-hosting.
Questions people ask
Is there a free plan?
Yes. Free costs nothing, requires no credit card, and allows 200 requests per day.
What does Pay as you go cost?
Pay as you go is $5 a month and applies a 5% markup on model cost.
Which platforms are available?
Requesty is offered through an API and a web interface.
Can it be self-hosted?
No. Requesty describes itself as a managed platform and does not offer self-hosting.
How are prompts and outputs retained?
On self-serve plans, logging is enabled by default and encrypted in the EU for up to 30 days. Logging can be disabled per key, and organization-wide zero data retention is available on written request.
What the maker publishes
- Agent tools
- The integrations page lists tools including Claude Code, Cline, Roo Code, VS Code, GitHub Copilot, Codex, n8n, LibreChat and OpenWebUI. source
- Agent tools
- The integrations page lists Claude Code, Claude Cowork, Cline, Roo Code, VS Code, GitHub Copilot, Codex, OpenCode, OpenClaw, Pi, Anthropic SDKs, DeepSeek, n8n, LibreChat and OpenWebUI. source
- Analytics
- Analytics attribute cost and latency by model, team, user, key and custom request metadata. source
- API compatibility
- The official Python and JavaScript OpenAI packages can be used by changing the base URL and API key, with no Requesty-specific SDK required. source
- API surface
- Documentation describes inference endpoints for chat, embeddings, audio, images and model listing plus management APIs for keys, groups, members and organization settings. source
- Caching
- Automatic prompt caching is offered, with the company claiming savings of up to 90% on repeated context. source
- CLI licence
- The published Requesty CLI repository is licensed under MIT. source
- Company founding
- LinkedIn lists Requesty as founded in 2023. source
- Company headquarters
- LinkedIn lists London, England as headquarters and company size of 2–10 employees. source
- Company legal name
- Crunchbase identifies the legal entity as Requesty Ltd. source
- Company register
- Dealroom lists company register number 15165717 and a 2023 launch date. source
- Compatibility
- Its API is OpenAI-compatible, and the site says existing SDK code can work by changing the base URL and API key. source
- Compliance
- The security page says its SOC 2 Type II programme is in progress and that it provides a GDPR Article 28 DPA on request. source
- Compliance status
- Requesty states that its SOC 2 Type II program is in progress, offers a GDPR Article 28 DPA on request and supports EU data residency. source
- Data retention
- Prompt and output logging is enabled by default for self-serve plans for up to 30 days in the EU, with per-key disablement and organization-wide zero-retention available on request. source
- Developer documentation
- The quickstart includes Python, TypeScript and cURL examples, response headers, request metadata, fallback policies and Anthropic SDK routing. source
- EU processing
- An EU gateway runs in Frankfurt on AWS eu-central-1 at router.eu.requesty.ai. source
- EU residency
- Requesty identifies Frankfurt as its EU gateway location and says EU requests can be routed there. source
- Fallback routing
- Policies can retry failed or timed-out requests on backup models without application retry code. source
- Funding
- Crunchbase lists one seed funding round with three investors, including The Twenty Minute VC and Tapestry VC, but obscures the amount. source
- Governance
- The site describes PII detection and scrubbing, content guardrails, team roles, model allowlists and audit logs. source
- Headquarters
- London, United Kingdom source
- Hosted deployment
- Requesty presents itself as a gateway service reached through router.requesty.ai and regional endpoints; no self-hosted server package is documented on the official pages reviewed. source
- Independent listing usage
- Appz lists Requesty as a web developer tool and reports zero people using it on that directory. source
Checked against the maker's pages on 21 Oct 2026.