Introduction
Hyphen is one key and one base URL in front of a small catalog of frontier-class models — under their real names.
Point an OpenAI- or Anthropic-compatible tool at the Hyphen gateway and drop in your key. That is the whole integration. One provider account, one bill, a short list of models, and a monthly cap you control.
Code
One base URL, one key
Every endpoint — Chat Completions, Responses, Anthropic Messages — is served from the same base URL and authenticated with the same key.
Three wire formats, one URL
Under the hood the gateway is an OpenAI-compatible proxy. It speaks three wire formats so your tools don't have to care what runs upstream:
| Endpoint | Wire format | Spoken by |
|---|---|---|
POST /v1/chat/completions | OpenAI Chat Completions | The default surface — OpenAI SDK, Cursor, Cline, Aider, Continue, Hermes |
POST /v1/responses | OpenAI Responses API | Codex CLI |
POST /v1/messages | Anthropic Messages API | Claude Code |
GET /v1/models | Model discovery | Any client |
Because the gateway translates between these formats, a tool that only speaks the Responses API (Codex) or only speaks Anthropic Messages (Claude Code) can still reach the same upstream models as everything else.
The models
Hyphen serves ten models — real provider model names, all from MiniMax. The ones most people start with:
| Model | Reach for it when… |
|---|---|
minimax-m3 | Flagship — hard reasoning, refactors, deep planning |
minimax-m2.7 | Balanced daily driver for day-to-day agent work |
minimax-m2.5 | Fast and cheap — high-volume, low-latency |
minimax-text-01 | Long-context — very large inputs |
The other six are latency-tuned -highspeed variants of the mid-tier models
(roughly twice the price of their standard siblings), the older minimax-m2.1
and minimax-m2 generations, and minimax-m2-her for role-play dialogue —
the full ten-row catalog is on the Models page.
These are the exact IDs the gateway accepts — no invented aliases; any other ID
returns an error. The catalog can change with notice, and during a provider
incident a request may be served by a smaller model of the same family — the
response's model field always reports what actually served it.
One key, one cap
Each subscription key carries a monthly budget. When it is spent, requests return HTTP 429 with the reset date — nothing runs over budget by accident, and the budget resets at the start of the next cycle. Prepaid credit funds a separate pay-as-you-go key you can switch to mid-month. See Rate limits & caps.
Status
Hyphen is new. Here is exactly what has been checked, so you can judge the rest of these docs accordingly.
Verified live against the gateway. The OpenAI Chat Completions surface, the OpenAI Responses surface, and the Anthropic Messages surface, including tool calling and streaming. Cap enforcement is verified too: a key past its budget returns HTTP 429.
Written from official docs. Every per-tool page in these docs is a setup guide built from that tool's own documentation. Treat it as a starting config, not a test report.
Scope today. One provider, MiniMax. There is no SLA. The gateway is best-effort and the Terms say so.
If a config here does not work, mail support@hyphen-solution.com. Bug reports go straight to the person who wrote the page.
Next steps
- Quickstart — get a key and make your first request in three steps.
- Models — the full model list and how to pick one.
- Choosing a model for which model fits which job, plus
the
max_tokensgotcha that returns empty responses. - Connect your coding agent — Claude Code, Codex CLI, Zed, Goose, Hermes, OpenClaw, and more.
- Chat UIs and automation for LibreChat, Open WebUI, Jan and n8n.
- Recipes for working code. Streaming, tool calling, structured JSON, agent loops, 429 handling, long context.
- API Reference — the full interactive spec.