wway
🇺🇸English 🇨🇳中文
Sign in Sign up
Multi-model gateway + usage console

One endpoint. Every model.

Collapse your Claude and OpenAI subscriptions into one standard endpoint. Protocols pass through untouched, native tools work as-is, and usage and cost live in one place.

Get started Open console
One unified endpoint Token-metered · official upstream rates Logs · usage · billing
Why wway

Not a translation shim — the upstream protocol reaches you unchanged.

Exact passthrough

Streaming, tool use, extended thinking, long context and prompt caching are forwarded with upstream semantics intact — message bodies and tool definitions are never rewritten.

Unified surface

One address and one key cover both the Anthropic Messages and OpenAI Responses protocols. Clients change a single environment variable.

Traffic visibility

Every call records its model, latency, tokens and cost. Query and export by model, time window or key for reconciliation.

Scheduling & failover

Accounts are scheduled automatically and a session stays pinned to the same upstream. When one is rate-limited, traffic moves over and context is rebuilt.

Channels

Two subscription channels covering the leading frontier models.

A
claude-max20
Messages API · anthropic
  • claude-opus-5
  • claude-sonnet-5
  • claude-fable-5-1
  • claude-opus-4-8
  • claude-haiku-4-5
Long context Extended thinking Tool use Vision
O
codex-max20
Responses API · openai
  • gpt-6-astra
  • gpt-5.6-luna
  • gpt-5.6-sol
  • gpt-5.6-terra
  • gpt-image-2
Reasoning Tool use Vision Image gen
+
More channels
On the roadmap
  • Added on demand
Ecosystem

Compatible with the clients and frameworks you already use.

CC
Claude Code
Anthropic CLI
CX
Codex CLI
OpenAI CLI
CD
Codex Desktop
OpenAI desktop
OC
OpenCode
Terminal agent
CS
Cherry Studio
Desktop client
CB
ChatBox
Cross-platform client
SDKs & frameworks OpenAI SDKAnthropic SDKVercel AI SDK LangChainLangGraphLlamaIndex LiteLLMcurl
Getting started

From sign-up to first call in four steps — no application code changes.

Register an account

Email sign-up, no approval queue.

Create an API key

Generate one on the API Keys page. Keys start with sk-.

Point at wway

Swap the base URL. Everything else stays.

Use as before

Native clients need no changes; streaming and tool calls pass straight through.

Claude Code 
export ANTHROPIC_BASE_URL=https://wway.ai
export ANTHROPIC_AUTH_TOKEN=sk-********
Codex CLI · ~/.codex/config.toml 
[model_providers.wway]
name = "wway"
base_url = "https://wway.ai"
wire_api = "responses"
Pricing

Token-precise billing with the channel multiplier stated upfront.

Official upstream rates · transparent multiplier · no hidden surcharge

How a charge is computed

cost = official_rate_per_token × tokens × channel_multiplier
e.g. claude-opus-5 · no cache
Input rate$5.00 / 1M
Output rate$25.00 / 1M
Channel multiplier× 1.3
One turn: 20K in, 2K out
Official cost$0.150
You pay$0.195
e.g. claude-opus-5 · with cache
Cache write · 5m$6.25 / 1M
Cache read$0.50 / 1M
Output rate$25.00 / 1M
Channel multiplier× 1.3
Reusing the same 20K context
You pay$0.078
  • Per-token rates mirror the upstream provider's published prices — no invented currency
  • Both channels currently run at 1.3×, visible before you create a key
  • Cache hits bill at the cache-read rate, so long sessions that reuse context cost noticeably less
  • No monthly fee, no minimum, no lock-in — pay only for what you use
  • Multiplier changes are announced in advance, never adjusted silently
  • Routing, scoped keys, request logs and usage analytics are all included
Get started →
FAQ

What teams ask before putting a gateway in front of production traffic.

What is wway?
A gateway that turns subscription AI quota into a standard API. It is a thin forwarding layer between your client and the upstream providers: one endpoint, one key, one bill — while requests and responses themselves stay untouched.
How is it different from calling providers directly?
At the protocol level there is no difference — bodies and response streams are forwarded verbatim. What changes is that you maintain one address and one key, and get a single cross-provider view of usage, logs and billing.
Will my requests be rewritten?
No. Account selection and metering happen at the routing layer only; message content, tool definitions and system prompts are never modified.
Are streaming and tool calls supported?
Yes. SSE streaming, tool use, extended thinking, long context and prompt caching all pass through with native semantics — the full capabilities of Claude Code and Codex CLI are intact.
Can sessions from different clients cross over?
No. The gateway pins each session by hash to one upstream account. If that account becomes unavailable, traffic switches over and context is rebuilt.
Any problem with very large request bodies?
Very large bodies (tens of MB) can hit an edge timeout when uploaded over a slow link. Enable your client's history compaction, or keep sessions shorter.

One endpoint. Every model.

Make your first call within minutes of signing up.

Start now →