Requesty
LLM gateway alternatives

LLM gateway alternatives, compared

Nine AI gateways and LLM routers, from managed APIs to open-source proxies you host. How each one runs, what it charges, and when it is the right pick, from the vendors' own pages.

Speak to founders
9 gatewaysSourced October 2026Ties shown as ties
three questions that narrow it down
9 gateways
01Must the gateway run inside your network?
yes LiteLLM, TrueFoundry, Kong, or Portkey and Helicone on Enterprise
no Any managed gateway
02Should the router choose the provider for you?
yes OpenRouter
no Requesty, LiteLLM or Portkey: you set the order
03Do you need limits per user and team, and EU, US or AP routing?
yes Requesty, on every plan
no Shortlist on price and observability
Trusted by teams at
Shopify
Amadeus
Chargebee
Demandbase
Pfizer
PWC
Capgemini
Sage
Siemens
Relevance AI
Appnovation
Shopify
Amadeus
Chargebee
Demandbase
Pfizer
PWC
Capgemini
Sage
Siemens
Relevance AI
Appnovation
At a glance

How each gateway runs and charges

The two things that rule most options in or out before features come into it.

GatewayHow you run itOpen sourcePricing
RequestyManagedNo5% on top of model cost, 0% with your own keys
OpenRouterManagedNo5.5% on credit purchases, 8% on Business; BYOK free to $25,000 a month, then 5%
LiteLLMSelf-hostedYesFree open source; Enterprise is custom pricing; plus your servers
PortkeyManaged or self-hostedYes, the gatewayFree to 10k logs a month, Production $49 a month, Enterprise custom
HeliconeManaged or self-hostedYesFree to 10k requests, Pro $79 a month, Team $799 a month, plus usage
Kong AI GatewayManaged or self-hostedKong Gateway core isNot published; through Kong Konnect
Cloudflare AI GatewayManagedNoAvailable on all Cloudflare plans
Vercel AI GatewayManagedNoNo markup on provider token prices, including BYOK
TrueFoundryManaged or self-hostedNoFree to 30k requests a month and 3 seats; paid plans on their pricing page

From each vendor's own pricing and product pages, October 2026. Model counts are left out because vendors count models, endpoints and providers differently.

The options

What each one is, and when it fits

One honest line on when to pick each, including the ones that are not us.

Requesty

Managed

An AI gateway for 600+ models with routing policies, spend limits, PII and secrets guardrails, and analytics. EU, US and AP gateways on every plan.

Good fit when

You want to write the routing yourself, cap spend per key, user and team, and keep traffic in a region, without hosting anything.

How Requesty works

OpenRouter

Managed

One API and one bill for models from many providers. OpenRouter picks the provider, steered by your preferences. EU and US routing on Business and Enterprise.

Good fit when

You would rather the router choose the provider for you, or need a model only it lists.

Requesty vs OpenRouter

LiteLLM

Self-hosted

An open-source LLM gateway for 100+ providers with virtual keys, budgets, RPM/TPM limits, guardrails and an admin UI. Runs on Docker, Helm or Terraform with Postgres.

Good fit when

The gateway has to run inside your network or air-gapped, and you have people to operate it.

Requesty vs LiteLLM

Portkey

Managed or self-hosted

An AI gateway priced by recorded logs, with fallbacks, load balancing, caching, budgets and guardrails. Its site now presents it as Prisma AIRS AI Gateway from Palo Alto Networks.

Good fit when

You want an open-source gateway with a hosted tier, and log-based pricing suits your volume.

Requesty vs Portkey

Helicone

Managed or self-hosted

An open-source LLM observability platform with gateway features: caching, rate limits and automatic fallbacks. On-prem deployment on Enterprise.

Good fit when

Observability is the main job and the gateway is a feature you need alongside it.

Requesty vs Helicone

Kong AI Gateway

Managed or self-hosted

Governance for LLM, MCP and agent traffic on the Kong platform: multi-provider routing, semantic caching, prompt guards, PII sanitization and token quotas.

Good fit when

You already run Kong for your APIs and want AI traffic under the same platform team.

Requesty vs Kong AI Gateway

Cloudflare AI Gateway

Managed

A proxy for AI requests on the Cloudflare network with caching, rate limiting, retries and model fallbacks, analytics and logs.

Good fit when

Your app already runs on Cloudflare Workers and you want visibility and caching in front of providers.

Requesty vs Cloudflare AI Gateway

Vercel AI Gateway

Managed

A hosted gateway callable from any infrastructure, with provider and model fallbacks, budgets per team, project, key or member, and request logs across text, image and speech models.

Good fit when

You build with the AI SDK or host on Vercel, and zero markup matters more than routing policies.

Requesty vs Vercel AI Gateway

TrueFoundry

Managed or self-hosted

An enterprise AI gateway that deploys to your VPC, on-prem or air-gapped, with RBAC, guardrails, rate limits and quotas, alongside self-hosted model serving.

Good fit when

You also serve your own models on Kubernetes and want the gateway in the same platform.

Choosing

Six things to compare

Feature lists look alike. These are the questions that separate gateways in production.

  1. 01
    Who runs it

    Managed means no servers, database or patches. Self-hosted means the gateway, and every provider key it holds, is yours to secure.

  2. 02
    Who picks the provider

    Some routers choose for you. Others let you name the provider and write the fallback order, so you know what served each request.

  3. 03
    How it charges

    A percentage on top of model cost, a fee on credit purchases, a price per logged request, a subscription, or zero markup with fewer controls.

  4. 04
    Where traffic goes

    Check which regions you can pin to, on which plan, and whether requests outside them are rejected.

  5. 05
    Spend and access control

    Limits per key are common. Limits per user and team, model allowlists and SSO are where gateways start to differ.

  6. 06
    What you can see

    Cost and latency per model are table stakes. Sessions, tool calls and usage per end customer take more.

Governance

Where Requesty differs: control per team

Decide which models and regions each team can use, mask PII before it reaches a provider, hand auditors the evidence, and sync your teams from Entra.

Groups
Allowed regions: EU, US
Group
Members
Access list
Region
Engineering
24
Engineering models · 18 models
EU
Support
32
Support models · 4 models
EU
Research
15
All approved · 42 models
EU+US
Legal
6
EU residency · 7 models
EU
Admin
Group admin
Member
Zero data retention: required

LLM gateway questions

It depends on how you want to run it. For a managed gateway where you write the routing and cap spend per user and team, Requesty. To self-host inside your network, LiteLLM. To let the router pick providers for you, OpenRouter. If you already run Kong or Cloudflare, their AI gateways fit that stack.

LiteLLM is fully open source and built to be self-hosted. Portkey and Helicone also publish open-source code, with hosted tiers on top. Open source means you run and secure it yourself, including the provider keys it holds.

Requesty, if you want to choose the provider and fallback order yourself, need EU, US or AP routing without a plan upgrade, or run most traffic on your own keys at 0%. Vercel AI Gateway if zero markup is the priority. LiteLLM if you need to self-host.

Self-host when the gateway must run inside your network or air-gapped and you have people to operate it. Otherwise a managed gateway removes the servers, the database and the patch cycle, and the gateway is no longer a box holding every provider key that you have to secure.

Pricing models differ: Requesty adds 5% to model cost (0% with your own keys), OpenRouter charges 5.5% on credit purchases, Portkey and Helicone charge by logged requests or subscription, Vercel adds no markup, and LiteLLM is free to run but costs servers and time.

Try a managed AI gateway

One base URL for 600+ models, with routing, limits and analytics. Free models need no credit card.

Speak to founders