Credits · one key · four APIs
The model gateway that fits the tools you already run.
Invite-only signup, mint an mmux_ key, and point Claude Code, Cursor, Codex, or Gemini at ModelMux. One catalog, metered usage, no upstream credentials on your laptop.
- Four APIs
- Anthropic Messages, OpenAI Responses, Chat Completions, Gemini generateContent
- Metered credits
- Holds, grants, and per-token rates in micros. Balance never leaves the server.
- Operated catalog
- Plans, RPM, rewrite rules, and upstream keys without a deploy.
Models
Public slugs and raw ids both work — claude-sonnet-5 or anthropic/claude-sonnet-5.
| Model | Id | Context | In / 1M | Out / 1M |
|---|---|---|---|---|
| GPT-6 AstraFeatured | openai/gpt-6-astra | 1M | $5 | $25 |
| Claude Opus 5 | anthropic/claude-opus-5 | 1M | $15 | $75 |
| Claude Sonnet 5 | anthropic/claude-sonnet-5 | 1M | $3 | $15 |
| Claude Fable 5.1 | anthropic/claude-fable-5.1 | 1M | $0.80 | $4 |
| GLM 5.3 | z-ai/glm-5.3 | 1M | $1 | $3 |
Four drop-in endpoints
POST https://api.modelmux.xyz/v1/messages
ANTHROPIC_BASE_URL=https://api.modelmux.xyz
ANTHROPIC_API_KEY=mmux_…
Origin only — do not append /v1.
POST https://api.modelmux.xyz/v1/chat/completions
base_url=https://api.modelmux.xyz/v1
model_provider = "modelmux"
base_url = "https://api.modelmux.xyz/v1"
wire_api = "responses"
Put model and model_provider above the table.
POST https://api.modelmux.xyz/v1beta/models/{model}:generateContent
x-goog-api-key: mmux_… or ?key=
For customers
- USD credits, tracked in micros. Holds on the way in, capture on the way out.
- Named keys. The secret is shown once, then only a prefix.
- Plan RPM and allowed models applied at the gateway.
- Usage by hour and model, with cost in credits.
Plans
Start free. Upgrade or grant credits from admin when you are ready.
Invite code required at signup
- 10 RPM
- 2 concurrent requests
- claude-sonnet-5 and glm-5.3
Included credits · full catalog
- 60 RPM
- 8 concurrent requests
- All models, including GPT-6 Astra
- Top up from the dashboard anytime