LLM Gateway Review
llmgateway.io · checked 15 Oct 2026
A single OpenAI-style gateway that sends traffic to more than 250 models across 40-plus providers.
How it plugs in
Signal on LLM gateways8.3#2 of 33
LLM Gateway puts one OpenAI-compatible API between applications and many model providers. It handles provider selection, fallback attempts, usage and spend measurement, response caching, and budget controls. The core gateway is AGPLv3 open source and can run on your own server. A free hosted plan is available, with enterprise options for larger teams.

Review
In a line
Hosted deployments and self-managed instances both let applications access more than 250 models from over 40 providers through LLM Gateway's OpenAI-compatible API.
What it does
Applications send requests through the gateway instead of connecting separately to each provider. You can change model identifiers without rewriting provider integrations.
The gateway may weigh availability, response speed, throughput, cost, and cache information when choosing a route. If a provider fails, times out, or rate-limits a request, the gateway can try another provider for that model up to two times.
Dashboards show requests, tokens, spending, latency, errors, and cache status by model, provider, project, or API key. Response caching can reduce repeated-request time and token costs.
How it connects
The service works through the API at api.llmgateway.io/v1 and supports existing OpenAI software libraries, cURL, any programming language, and a dedicated AI SDK package. The API reference covers chat completions, responses, Anthropic messages, embeddings, images, speech, transcription, moderation, search, video, and realtime operations.
Guides cover Claude Code, Cursor, Codex CLI, Cline, Continue CLI, GitHub Copilot, n8n, OpenCode, and VS Code. The gateway, dashboard, API, and worker use AGPLv3, while code in the ee directory is commercially licensed.
You can bring your own provider keys with no platform fee and pay providers directly. The product runs as an API, browser extension, Linux software, web service, or self-hosted deployment.
Good fit
The product is aimed at product and platform teams seeking a shared connection to numerous model providers. It also fits organizations deploying shared model infrastructure across teams.
Spending limits and optional alerts are available for API keys and members. Enterprise plans add team budgets, organization and project roles, SSO, SCIM provisioning, guardrails, audit logs, and contractual service levels.
Watch for
Self-hosting the core gateway is free, but enterprise code requires a license. Hosted credit purchases use provider list token rates plus a flat 5% fee; international cards add 1.5%.
On the Free plan, free-model requests are limited to 20 per minute, and logs retain metadata rather than full payloads. Full payload retention costs $0.01 per 1 million tokens, while Enterprise includes 30 days of full payload retention and unlimited data retention.
The company says it is SOC 2 Type II audited and GDPR-aligned, while its trust center lists GDPR as in progress. Check the retention, licensing, and compliance terms that apply to your deployment.
Pricing
The Free plan costs nothing indefinitely. It offers 200-plus models, over 40 providers, rate-limited free models, metadata-only logs, activity-log export, and Discord Community support.
Enterprise pricing is price on request and billed on a custom basis. It comes with a 30-day trial, all models, custom routing, 30 days of full payload retention, SSO with SAML or OIDC and SCIM, audit logs, guardrails, a contractual SLA, and 24/7 support.
Hosted pay-as-you-go credits add a 5% fee to provider list token rates, plus 1.5% for international cards. Bringing your own provider keys carries no platform fee.
Verdict
Choose LLM Gateway if your applications need a single OpenAI-compatible connection to many providers, with fallback routing and detailed usage controls. Its open-source core also gives teams a self-hosting option.
The Free plan is suitable for trying the service or using provider keys without a platform fee. Larger organizations should review the commercial licensing, retention charges, enterprise identity features, and support terms before moving shared workloads onto it.
Our take
LLM Gateway suits teams that need one connection for switching among many model providers instead of maintaining separate application integrations. Its retry behavior, dashboards, caching, and spending controls cover common operating tasks, while AGPLv3 code allows self-hosting. The free plan is limited by rate caps and metadata-only logs, and hosted credits add a 5% fee. Enterprise buyers get broader identity, retention, guardrail, support, and service-level options, but must request a custom price and account for commercial licensing.
- A unified OpenAI-style endpoint reaches 250-plus models from more than 40 providers
- Provider failures, timeouts, and rate limits can trigger two further attempts
- Dashboards cover usage, latency, errors, caching, and spend
- The AGPLv3 core is available for free self-hosted deployment
- Using your own provider keys carries no platform fee
- Free-model traffic on the Free plan stops at 20 requests per minute
- Full payload retention is charged at $0.01 per 1 million tokens
- Buying hosted credits incurs a flat 5% fee
- Enterprise code requires a commercial license
- Enterprise pricing is available only on request
- Pick it ifProduct and platform teams sharing access to many model providers; Organizations needing routing, fallback, usage tracking, and budget controls; Teams that want an open-source self-hosting option
- Skip it ifYou need unrestricted free-model traffic on the Free plan; You require full payload logs without an added retention charge; You want published Enterprise pricing instead of a custom quote
Pricing
The no-cost plan remains available indefinitely, and Enterprise offers a 30-day trial. Purchased hosted credits follow provider list token rates with an added 5% fee; international cards add 1.5%. BYOK has no platform fee, while full payload retention is charged at $0.01 per 1 million tokens.
- Freeforever · 200+ models · 40+ providers · free models rate-limited to 20 requests/minute · metadata-only logs · Discord supportFree
- Enterprisecustom · 30-day trial · all models · custom routing · full payload retention for 30 days · SSO/SAML/OIDC with SCIM · audit logs · guardrails · contractual SLA · 24/7 supportOn request
As llmgateway.io lists them, checked 15 Oct 2026.
Questions
What is LLM Gateway?
It can be hosted or self-managed, giving applications one OpenAI-compatible endpoint for over 250 models supplied by more than 40 providers.
Can I use my own provider keys?
Yes. BYOK has no platform fee, and you pay the model providers directly.
Does LLM Gateway use an open-source license?
The gateway core, dashboard, API, and worker use the AGPLv3 license. Files under the ee directory are commercially licensed.
Does it retry failed requests?
Yes. Following an error, timeout, or rate limit, the gateway may try another provider for that model, up to two additional times.
What is included with the Free plan?
The plan remains free and offers 200-plus models, more than 40 providers, metadata-only logs, activity-log export, and Discord Community support. Free models are capped at 20 requests per minute.
How much do hosted credits cost?
Purchased credits follow provider list token rates and include a flat 5% charge; international cards incur another 1.5%.
What does Enterprise add?
Enterprise offers all models, custom routing, identity features, audit logs, guardrails, 30 days of complete-payload retention, a contractual SLA, and round-the-clock support. Pricing is on request.