Every model.
One key.
Honest billing.
A drop-in Anthropic & OpenAI-compatible gateway for Claude Code. Opus 4.8 for the hard stuff, Sonnet 4.6 for daily work, Haiku 4.5 for speed. Same models, same SDK — every token 30% cheaper than going direct.
Every model. One key.
Opus 4.8 for the hard stuff, Sonnet 4.6 as your daily driver, Haiku 4.5 when speed and cost win. Every rate comes in 30% under what the lab charges direct.
Claude Opus 4.8
Top-tier reasoning, complex coding, long-running agents.
- Multi-step refactors across large codebases
- Architectural planning & deep research
- Highest-quality long-form writing
Point your SDK at api.anvat.app — done.
Keep the same Anthropic or OpenAI SDK. Same streaming, same tool calls, same /v1/messages. Just swap the base URL and pay materially less.
# Drop-in for Claude Code
export ANTHROPIC_BASE_URL="https://api.anvat.app"
export ANTHROPIC_API_KEY="sk-anvat-..."
# Or with the OpenAI SDK
client = OpenAI(
base_url="https://api.anvat.app/v1",
api_key="sk-anvat-..."
)
# Same models. Same SDK. 30% less.
→ claude-opus-4-8
→ claude-sonnet-4-6
→ claude-haiku-4-5Frontier models at half the effective cost.
30% below list
Metered usage billed 30% under Anthropic and OpenAI's published rates. No markup, no rounding.
Credits that double
Prepaid packs double your spend — pay $100, receive $200 in API credit on every tier.
Drop-in gateway
Set ANTHROPIC_BASE_URL or your OpenAI base URL. Keep your SDK, your streaming, your tool calls.
Honest activity logs
Transparent per-token billing. See every request, every model, every cost — no fine print.
No prompt storage
We don't store prompts. We don't train on your data. Just routing and metering.
One key, every model
Opus, Sonnet, Haiku today. GPT-5.5, Gemini 3.1, Grok 4.3, Kimi K2.5 rolling out next.
Built by engineers who run agents in production — for engineers who do the same.
Anvat is an OpenAI and Anthropic compatible API gateway built for developers who ship with Claude Code, Cursor, Codex, and custom agents. Point ANTHROPIC_BASE_URL or your OpenAI SDK base URL at api.anvat.app — keep the same models, the same streaming, the same tool calls. Pay materially less.
Metered usage is billed at 30% below published list rates from Anthropic and OpenAI. Prepaid credit packs double your spend on every tier — pay $100, receive $200 in API credit. Most teams see roughly half the effective cost versus routing the same Opus 4.8, Sonnet 4.6, and Haiku 4.5 workloads direct to providers at full price.
Anvat is a drop-in LLM gateway and cheap Claude API alternative to juggling provider keys, brittle proxies, or opaque consumer plans. One key unlocks every frontier model with honest per-token billing, transparent activity logs, and subscriptions that undercut comparable Claude consumer plans. No fine print. No prompt storage. No training on your data.
- Honest billingPer-token rates published on the page. What you see is what you pay.
- Zero data retentionPrompts and completions are never stored or used for training.
- Same SDKsAnthropic & OpenAI SDKs, /v1/messages, streaming, tool use — unchanged.
- Backed by uptimeMulti-region routing with provider-side failover and 99.9% target.
"Switched our Claude Code fleet to Anvat in five minutes. Bill went from painful to predictable — same Opus, same Sonnet, no code changes."
"The honest part isn't marketing. Cache reads price out exactly as advertised, and the activity log lines up to the token."
"We run multi-agent pipelines on Sonnet and Haiku all day. Anvat shaved ~45% off our monthly LLM spend with credits doubled in."
Frequently asked questions.
Everything teams ask before they swap their base URL.
What is Anvat?＋
Anvat is a drop-in Anthropic & OpenAI-compatible LLM gateway built for Claude Code, Cursor, Codex, and custom agents. It routes the same frontier models (Claude Opus 4.8, Sonnet 4.6, Haiku 4.5) through one API key at 30% below the lab's published list rates — no markup, no rounding, no fine print.
How does Anvat make Claude 30% cheaper?＋
We operate at scale on negotiated infrastructure and pass the savings on. Metered usage is billed exactly 30% under Anthropic's and OpenAI's public per-token rates. Prepaid credit packs also double on every tier — pay $100, receive $200 in API credit — bringing the effective cost to roughly half of going direct.
Is Anvat really a drop-in replacement?＋
Yes. Point ANTHROPIC_BASE_URL at https://api.anvat.app or set the OpenAI SDK base_url to https://api.anvat.app/v1. Keep the same SDK, the same streaming, the same tool calls, the same /v1/messages contract. Nothing else changes in your code.
Do you store or train on my prompts?＋
No. Anvat does not store prompts and does not train on customer data. We route requests, meter tokens, and log activity for billing transparency — that's it.
Which models are supported?＋
Live today: Claude Opus 4.8, Claude Sonnet 4.6, Claude Haiku 4.5. Rolling out next on the same endpoint: GPT-5.5, Gemini 3.1 Pro, Grok 4.3, and Kimi K2.5 — every model 30% under list price.
Who is Anvat for?＋
Developers and teams running Claude Code, Cursor, Codex, and custom agents who want frontier-model quality without paying full provider list prices — and who want one key, honest per-token billing, and transparent activity logs instead of opaque consumer subscriptions.
One key. Every model.
Start in 60 seconds.
Sign up, drop in your base URL, and ship. $2 free credit on every new account — no card required.