LLM Gateway
An open-source API gateway that routes requests across 200+ LLMs from 40+ providers through a single OpenAI-compatible endpoint, with real-time cost tracking and analytics.
At a Glance
About LLM Gateway
LLM Gateway is an open-source API gateway for large language models, built by theopenco and licensed under AGPLv3 for its core layer. It lets developers route requests to 40+ providers—including OpenAI, Anthropic, Google, AWS Bedrock, Azure, DeepSeek, and more—through a single unified API endpoint that is drop-in compatible with the OpenAI SDK. The project has accumulated over 1,400 GitHub stars since its April 2025 launch and is SOC 2 Type II certified.
What It Is
LLM Gateway sits as a middleware layer between your application and any LLM provider. Instead of managing separate API keys, dashboards, and SDKs for each provider, you point your existing OpenAI SDK at LLM Gateway's base URL and it handles routing, failover, cost tracking, and analytics automatically. The gateway supports 200+ models across text generation, embeddings, vision, reasoning, image generation, video generation, and more.
How the Routing Layer Works
The core value proposition is intelligent, automatic routing with built-in reliability:
- Automatic failover: When a provider goes down, traffic is rerouted to a healthy alternative in real time. The site claims that while individual providers average ~94% uptime (~22 days of downtime per year), combining multiple providers through automatic failover yields effective uptime of 99.9999%.
- Bring Your Own Keys (BYOK): Users can supply their own provider API keys and pay providers directly, with LLM Gateway providing routing, tracking, and analytics at no platform cost.
- Per-project routing overrides: Enterprise users can pin production workloads to specific regions or providers, set fallback chains, and apply cost ceilings per project.
- Prompt caching: Supported to reduce latency and cost on repeated prompts.
Open-Source Deployment Model
The core gateway is AGPLv3-licensed and self-hostable via Docker or Terraform modules. The repository's folder structure includes a Next.js dashboard (apps/ui), a playground (apps/playground), a Hono backend (apps/api), and the gateway routing layer (apps/gateway). Enterprise features—advanced billing, extended data retention, SSO, white-labeling, and team management—live in the ee/ directory under a separate commercial license. Self-hosting is free forever; the cloud-hosted version charges a 5% flat fee on credit top-ups, with no markup on token prices.
Observability and Cost Analytics
Every request through LLM Gateway is tracked with cost, latency, and token usage. The dashboard provides:
- Requests, tokens, total spend, and average cost per 1K tokens over 7 or 30 days
- Per-model and per-provider breakdowns to identify expensive outliers
- Error rate, cache hit rate, and reliability trend monitoring
- Project-level usage explorer with dedicated charts and tables
- Organization-wide analytics for enterprise deployments, rolling up spend across teams
Enterprise Capabilities
The enterprise tier adds a set of compliance and multi-team controls built for production deployments:
- SSO/SAML: SAML 2.0 and OIDC with SCIM provisioning and group-based role mapping
- Audit logs: Tamper-evident, SIEM-ready trails for every config change, key rotation, and admin action
- Guardrails: Server-side detection for prompt injection, PII, secrets, and policy violations, enforced at the gateway layer
- Provider compliance policies: Route only to providers meeting specified certifications (SOC 2, ISO 27001, GDPR, no prompt training)
- White-label chat playground: Deployable under a custom domain with custom branding
- Terraform modules: One-command provisioning on AWS, GCP, or Azure
Update: v1.8.0
The latest release is v1.8.0, published July 13, 2026, with the repository last updated July 15, 2026. The project has been actively developed since its April 2025 creation, with 25+ contributors and 157 forks on GitHub. Recent additions highlighted on the site include DevPass (fixed-price monthly plans for coding tools like Claude Code and Cursor), an MCP server, agent skills registry, and an AI SDK provider integration. The SOC 2 Type II certification was announced as a recent milestone.
Community Discussions
Be the first to start a conversation about LLM Gateway
Share your experience with LLM Gateway, ask questions, or help others learn from your insights.
Pricing
Free
Bring your own provider API keys and pay providers directly. Routing, tracking, and analytics included at no platform cost.
- All 200+ models
- Bring Your Own Keys (BYOK)
- 3 free rate-limited models
- Chat and API access
- Activity logs and export
Credits
Pay-as-you-go credits for any model at provider rates, with a 5% flat platform fee on top-ups. No subscription, no token markup.
- All 200+ models
- 5% flat fee on credit top-ups
- Pay provider rates directly
- No subscription required
- No token markup
- Full routing and analytics
Enterprise
Fully managed or self-hosted enterprise deployment with custom SLAs, SSO, audit logs, guardrails, and dedicated support.
- Everything in Free
- Volume-discounted platform fees
- All 200+ models
- Custom routing rules
- 3 free models with custom limits
- Unlimited data retention
- Enterprise audit logs
- Enterprise guardrails
- Per-project routing overrides
- Provider compliance policies
- Discord and Slack alerts
- SSO/SAML with SCIM provisioning
- White-label chat playground
- Admin controls
- Unlimited seats
- Custom rate limits
- 99.9% SLA guarantee
- 24/7 priority support via Slack/Discord
- Onboarding assistance
- Terraform modules for AWS, GCP, Azure
- 30-day production pilot
- Invoicing options
Capabilities
Key Features
- Unified OpenAI-compatible API endpoint
- 40+ LLM provider support
- 200+ models
- Automatic failover and reliability routing
- Bring Your Own Keys (BYOK)
- Real-time cost and token tracking
- Per-model and per-provider analytics
- Prompt caching
- LLM guardrails (prompt injection, PII, secrets detection)
- Enterprise audit logs
- SSO/SAML with SCIM provisioning
- Per-project routing overrides
- White-label chat playground
- Self-hosted deployment via Docker
- Terraform modules for AWS, GCP, Azure
- MCP server
- Chat playground
- Image and video generation studio
- Provider compliance policies
- Per-member budgets and spend controls
- Organization-wide analytics
- Discord and Slack alerts
