Hyper

Hyper is fast, cost-effective inference built for agentic coding. Grab a key, point your tool at it, and start shipping. Most people are up and running in under two minutes.

Why Hyper

Fast and cost-effective. We partner closely with cloud and silicon providers to optimize performance at the hardware and kernel level, so you get quick responses at a fraction of the price.
Models exactly as they were built. No quantized or trimmed-down variants — original reasoning options and behaviors are preserved.
Built for teams. Master keys, project-specific sub-keys, and pre-paid credits give you governance over how your organization spends.
Optional ZDR. Pay As You Go (PAYG) and Business & Enterprise subscriptions are ZDR (Zero Data Retention). We only use data from Personal subscriptions to train new models, but we still anonymize your data to protect your privacy.

Every model in the catalog is hand-picked for software development — vetted for code generation, debugging, refactoring, and multi-file editing.

Get your key

Head to hyper.charm.land.
Sign up with Google, GitHub, or email.
Create a new API key; Click "New key" to open the API key configuration modal.
Configure it as you wish!
Now you are ready to go, just copy the key.

HOLD ON!

Make sure you save it somewhere, you will not be able to copy it again 😄

Copy API key

Keep it handy

Export it once and every Hyper-aware tool will pick it up:

bash

export HYPER_API_KEY="sk-hyper-xxxxxxxxxxxxxxxxxxxxxxxx"

Connect your coding agent

Pick your tool and follow the two-minute setup. If you're not sure where to start, Crush has first-class Hyper support.

Crush

First-class Hyper support — sign in during first run and it just works.

Access Hyper's models in Pi via the OpenAI-compatible endpoints.

Zed

Use Zed's Agent or Terminal Threads through Hyper.

Claude Code

Route Claude Code through Hyper's Anthropic-compatible API.

Codex

Point Codex at Hyper's OpenAI-compatible endpoints.

OpenCode

Have OpenCode talk to Hyper's OpenAI-compatible endpoints.

Or call the API directly

Hyper speaks OpenAI and Anthropic out of the box. Swap the base URL, drop in your key, and go:

cURLPythonNode.jsGo

bash

curl https://hyper.charm.land/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "model": "deepseek-v4-flash",
    "messages": [
      {"role": "user", "content": "Hello!"}
    ]
  }'

python

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_API_KEY",
    base_url="https://hyper.charm.land/v1"
)

response = client.chat.completions.create(
    model="deepseek-v4-flash",
    messages=[{"role": "user", "content": "Hello!"}]
)

print(response.choices[0].message.content)

import OpenAI from "openai";

const client = new OpenAI({
    apiKey: "YOUR_API_KEY",
    baseURL: "https://hyper.charm.land/v1",
});

const response = await client.chat.completions.create({
    model: "deepseek-v4-flash",
    messages: [{ role: "user", content: "Hello!" }],
});

console.log(response.choices[0].message.content);

package main

import (
	"context"
	"fmt"
	"os"

	"charm.land/fantasy"
	"charm.land/fantasy/providers/openaicompat"
)

func main() {
	provider, _ := openaicompat.New(
		openaicompat.WithBaseURL("https://hyper.charm.land/v1"),
		openaicompat.WithAPIKey("YOUR_API_KEY"),
	)

	ctx := context.Background()
	model, _ := provider.LanguageModel(ctx, "deepseek-v4-flash")

	agent := fantasy.NewAgent(model)
	result, err := agent.Generate(ctx, fantasy.AgentCall{Prompt: "Hello!"})
	if err != nil {
		fmt.Fprintln(os.Stderr, err)
		os.Exit(1)
	}
	fmt.Println(result.Response.Content.Text())
}