A curated list of AI API gateways, proxy services, crypto-payment LLM platforms, and self-hosted alternatives. Maintained by Stormfire.
An AI API gateway is a service that sits between your application and one or more AI providers (OpenAI, Anthropic, Google, etc.), giving you a single endpoint for many models. Common reasons to use one:
- Unified billing across multiple providers
- Failover and load balancing between models or providers
- Geographic access when your country's payment infrastructure can't reach the upstream
- Alternative payment methods (crypto, prepaid, regional methods)
- Caching and rate limiting at the gateway layer
- Observability — logs, metrics, cost attribution per app or user
This list catalogs services we believe are worth comparing, both hosted and self-hosted.
- Hosted Gateways — Crypto Payment
- Hosted Gateways — Traditional (Card/Stripe)
- Hosted Gateways — Open-Source Model Focus
- Self-Hosted Gateways
- Free-Tier Aggregator Gateways
- Direct Providers (for comparison)
- Pricing & Latency Comparison Tools
- Articles & Reading
- Contributing
- License
Services that accept cryptocurrency as the primary or only payment method. Best for developers in regions where Stripe / cards face friction.
| Service | Models | Platform Fee | Crypto Accepted | KYC | Min. Top-up |
|---|---|---|---|---|---|
| Stormfire | 30+ | 0% | USDT (TRC20/ERC20/BEP20/Polygon) | None | $15 USDT |
| OpenRouter | 300+ | 5.5% | USDC only | $10 USDC | |
| MixRoute | 200+ | varies | USDT + USDC | None | $10 |
ℹ️ Stormfire maintains this list. We've done our best to be accurate and fair. PRs welcome to correct numbers, add competitors, or update offerings.
| Service | Models | Pricing Model | Special Features |
|---|---|---|---|
| OpenRouter | 300+ | Pass-through + 5.5% | Largest model catalog, OAuth flows |
| Anyscale Endpoints | Open-source | Pay-per-token | Self-hosted upgrade path |
| Together AI | Open-source | Pay-per-token | High-throughput inference |
| Replicate | Multi-modal + LLMs | Pay-per-second | Image / video focus, custom models |
| Fireworks AI | Open + frontier | Pay-per-token | Low-latency inference, function calling |
| Groq | Llama / Mixtral | Pay-per-token | Hardware-accelerated inference |
| Perplexity API | Online + Llama | Pay-per-token | Built-in web search |
| DeepInfra | Open-source | Pay-per-token | Cheap inference, no minimums |
| TogetherAI | Open-source | Pay-per-token | Compatible with OpenAI SDK |
| Service | Models | Notable |
|---|---|---|
| Hugging Face Inference Endpoints | All HF models | Closest to "Hugging Face hosted" |
| Modal | Custom containers | Better for inference platforms than pure API |
| Banana | Custom models | Container-first, GPU autoscale |
| Replicate | Public + private models | Strongest community library |
| RunPod Serverless | Custom | GPU-on-demand with cold-start optimizations |
Run your own gateway on your infrastructure. Best for orgs that need full data control, on-premise inference, or want to wrap multiple providers behind one API.
| Project | Stack | Key Features | Stars |
|---|---|---|---|
| LiteLLM Proxy | Python / FastAPI | 100+ providers, OpenAI-compatible, observability | 14k+ |
| New API | Go + React | Multi-tenant, billing, admin UI, OneAPI fork | 8k+ |
| One API | Go + Vue | Original fork base, simpler, very mature | 18k+ |
| Portkey Gateway | Node.js | Edge-ready, observability, fallback logic | 6k+ |
| Helicone | TypeScript | Observability layer, supports any upstream | 3k+ |
| LangFuse | TypeScript | Tracing + analytics, can sit alongside gateway | 6k+ |
| Kong AI Gateway | Lua / OpenResty | Enterprise gateway plugin for AI | n/a |
Self-hosted projects that work by stitching together free tiers from multiple providers. Different from traditional self-hosted gateways: the value proposition is "don't pay, route through free quotas" rather than "unify billing across paid providers."
Best for hobby projects, learning, and personal use. Not suitable for production — free tiers have unpredictable availability, no SLA, and ToS that may prohibit aggregation.
| Project | Stack | Approach | Notable |
|---|---|---|---|
| OmniRoute | TypeScript | 160+ providers / 50+ free sources; RTK+Caveman compression saves 15-95% tokens | Routes Claude Code / Codex / Cursor / Cline / Copilot through free Claude/GPT/Gemini tiers; Desktop + PWA |
| freellmapi | Python | Aggregates 16 free-tier providers behind one OpenAI-compatible /v1 endpoint |
~1.7B tokens/month aggregate free budget; minimal config |
| GPT4Free | Python | Reverse-engineers public chat UIs to expose a free API | Long-running flagship of the category; cat-and-mouse with providers |
⚠️ Note on ToS: Most free-tier aggregators sit in legally gray territory. Provider ToS typically prohibits programmatic aggregation of personal free quotas for redistribution. Use at your own risk for personal / learning purposes; do not rely on these for paying customers.
For reference — the upstream providers themselves.
| Provider | Models | Billing | Restrictions |
|---|---|---|---|
| OpenAI | GPT-4o, GPT-5, o1, o3 | Card via Stripe | Stripe risk model rejects 30+ countries |
| Anthropic | Claude 3.5, Claude 3.7 | Card via Stripe | Same Stripe issues |
| Google AI Studio | Gemini 2 Pro, Flash | Google Cloud Billing | Free tier, then Cloud Billing |
| DeepSeek | DeepSeek v3, v4 | Card / Alipay | China-based, but accepts intl cards |
| Mistral AI | Mistral Large, Small | Card | Mostly EU-friendly |
| xAI | Grok-3 | Card | US-leaning, limited geos |
| Cohere | Command R+ | Card | Enterprise-friendly |
| Tool | What it does |
|---|---|
| Stormfire Compare | Side-by-side: Stormfire vs OpenRouter vs MixRoute |
| Vellum AI Leaderboard | Model performance + cost benchmarks |
| Artificial Analysis | Latency + cost + quality charts |
| LLM Pricing Calculator | Multi-provider cost estimator |
| Helicone Cost Calculator | Real-time LLM cost calculator |
Foundational and recent writing on AI API access, gateway design, and crypto payment for AI:
- Why I Stopped Paying for AI APIs with My Credit Card — The Stripe risk model and AI access. (Stormfire blog)
- OpenRouter Alternative for Crypto Users — Side-by-side comparison and migration guide. (Stormfire blog)
- How to Pay for ChatGPT API with USDT — Step-by-step guide. (Stormfire blog)
- Awesome Crypto AI — Tangent list on AI x Crypto. (Community)
If you've written a high-quality, technically substantive article on these topics, open a PR.
Pull requests welcome. Please follow these rules to keep the list high-signal:
- One service per PR for easier review.
- No paid promotion. Entries are evaluated on technical merit only.
- Disclose affiliations in the PR description if you work for the service. We'll still review fairly, but transparency matters.
- Working URLs. Broken links will be flagged and removed in periodic sweeps.
- Be honest. Mention weaknesses where they exist (high fees, limited geo, slow inference). This list is meant to help readers decide.
- Markdown table format. Follow existing column conventions.
- No SEO spam. Listings that exist primarily to promote a service without adding catalogue value will be rejected.
Entries are removed when they:
- Become inactive (4+ weeks of downtime, abandoned site)
- Are found to be scam or malicious (rug-pulled, phishing, credential theft)
- Materially misrepresent their offering (e.g., claim "no fees" but charge hidden fees)
- Stop maintaining the underlying service
CC0 1.0 Universal (Public Domain). Copy, modify, and republish freely. Attribution appreciated but not required.