LLM gateway · Agent kits · Free tools

One API key for Claude, GPT, Grok and DeepSeek.

An OpenAI- and Anthropic-compatible endpoint with prepaid credits and published per-token rates, from $0.24 per 1M input tokens (DeepSeek V4 Flash).

Get API key Read the quickstart

First top-up: From $3
Billing: Pay per token; plans opening soon
Formats: /v1/chat/completions, /v1/messages

Request

curl https://api.oneaiapikey.com/v1/chat/completions \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-5",
    "stream": true,
    "stream_options": { "include_usage": true },
    "messages": [{
      "role": "user",
      "content": "Name the HTTP 429 status."
    }]
  }'

200 OKtext/event-stream · abridged

data: {"choices":[{"delta":{"content":"Too"}}]}data: {"choices":[{"delta":{"content":" Many"}}]}data: {"choices":[{"delta":{"content":" Requests."}}]}data: {"choices":[],"usage":{"prompt_tokens":15,"completion_tokens":4}}data: [DONE]

15 in × $2.2/1M + 4 out × $11/1Mbilled $0.000078

Example call to claude-sonnet-5. The cost line uses the published rate for that model.

Works with

Plus any SDK or app that accepts a custom OpenAI or Anthropic base URL.8 setup guides

Request flow

What happens between your code and the model.

Every call takes the same six steps. Nothing is hidden from you: the key rules you set apply before the provider is reached, and the cost of each request is in your log as soon as it finishes.

Your app
Sends an OpenAI- or Anthropic-format request to api.oneaiapikey.com with your key. Only the base URL and key change in your code.
Key check
The key must be active and unexpired, allowed to call the requested model, and inside its spend limit. Your balance must cover the call.
Routing
The model name selects the upstream channel that serves it. You never configure a provider account yourself.
Provider
The request is forwarded to the model provider, and tokens stream back as they are generated.
Metered response
Input, output and cached tokens are counted, priced at the published rate and deducted from your prepaid credit.
Usage log
Each request writes one row: time, key, model, tokens and cost. You see it in the console, per key.

Products

Gateway, agent kits and tools on one account.

The gateway is the paid part. The agent kits and tools are free, and exist so the gateway is easy to use well.

01 · GATEWAY

One endpoint, a key per project.

OpenAI- and Anthropic-compatible. Give every project or agent its own key with a spend limit and a model allow-list, then see exactly what each key cost.

API keysExample data

Name	Key	Models	Spent / limit	Status
production	sk-…7f2a	All models	$12.40 / $50.00	Active
ci-review	sk-…c91e	claude-haiku-4-5	$1.85 / $5.00	Active
old-laptop	sk-…04bd	All models	$3.10 / None	Disabled

OpenAI format: https://api.oneaiapikey.com/v1/chat/completions
Anthropic format: https://api.oneaiapikey.com/v1/messages
Model list: https://api.oneaiapikey.com/v1/models

How the gateway works

02 · AGENT KITS

Coding agents, already configured.

Step-by-step setups for Claude Code, Codex CLI, Cursor and 5 more. Two variables, then your agent runs on your balance.

~/.zshrc

# Claude Code, in your shell profile
export ANTHROPIC_BASE_URL=https://api.oneaiapikey.com
export ANTHROPIC_AUTH_TOKEN=sk-your-key

Browse agent kits

03 · FREE TOOLS

Estimate before you spend.

Cost calculator, agent config generator and token estimator. They run in your browser; no account needed.

Open the tools

Model: claude-sonnet-5
Input: 2.0M tokens
Output: 0.4M tokens
Estimate: $8.80

Chat

Talk to any model in the browser, on the balance your keys use.

The same catalogue as the API, without writing code. Pick a model per conversation; replies stream in as they are generated.

Model picker: Every model your account can call, with its per-token rate and a Free or Discount label where one applies.
System prompts: Presets for coding, short answers and editing, or your own instructions per conversation.
Your history: Conversations stay in your browser. Search them, rename them, export one as a file.
One balance: Chat draws on the same balance as your API keys, and every reply shows up in your usage log.

Open chat How chat works

New chat

Recent

Process vs thread
SQL or NoSQL for event logs
Meeting confirmation email

DeepSeek V4 Flash$0.24 / $0.96 per 1MExample

What is the difference between a process and a thread? Two sentences.

A process is a running program with its own memory, isolated from other processes by the operating system. A thread runs inside a process and shares that memory with its sibling threads, so it is cheaper to start but needs care when two threads change the same data.

Now a Python retry with exponential backoff.

Wait twice as long after each failed attempt:

python Copy

for attempt in range(5):
    try:
        return send()
    except RateLimitError:
        time.sleep(2 ** attempt)

112 in · 54 out$0.000078 Copy Regenerate

Message DeepSeek V4 Flash

Enter to send, Shift+Enter for a new line

Product preview with an example conversation. Token counts are illustrative; the rate is DeepSeek V4 Flash's published price.

Pricing

Published rates, billed per token.

No seats and no monthly minimum. Each request costs the tokens it used at the rate shown here, paid from your wallet.

USD per 1M tokens for 5 of 11 models
Model	ID	Input	Output	Cached
Claude Opus 5.5	claude-opus-5-5	$4.40	$22	$0.22
Claude Sonnet 5	claude-sonnet-5	$2.20	$11	$0.22
GPT-6.1 Sol	gpt-6.1-sol	$2.20	$11	$0.11
Grok 4.6	grok-4.6	$2.20	$6.60	$0.55
DeepSeek V4 Flash	deepseek-v4-flash	$0.24	$0.96	$0.0048

USD per 1M tokens. 11 models in total, each with its own page.

Pricing, plans and credit packs All models and rates

Ways to pay

Pay as you go$0 / mo

Top up a prepaid wallet with $3 or more, or a credit pack from $10. One credit is $1 of usage. Every model, nothing renews.

Monthly plansOpening soon

A weekly allowance that pays for any model, measured in DeepSeek V4 Flash input tokens. Plus and Max also take a discount off every model. Until plans open, every account pays as it goes.

Pro3.4M DeepSeek V4 Flash input tokens a week (up to 13M with cache hits)$5
Plus6.4M DeepSeek V4 Flash input tokens a week (up to 25M with cache hits), 2% off every model$9
Max21.2M DeepSeek V4 Flash input tokens a week (up to 81M with cache hits), 4% off every model$29

Plan prices are per month. "Up to" assumes about 76% of input is cache hits; output counts 4 times.

No free or discounted models right now. Rates on every model are the published list price.

Browse all models

Console

Your balance, your keys and every request, on one screen.

Product previewExample data

oneaiapikey.com/console

Relaykit

Overview
Keys
Usage
Billing
Settings

Balance

$33.23

Spent, last 7 days

$10.52

Active keys

Keys

production$12.40 / $50.00

ci-review$1.85 / $5.00

Recent requests

Time	Model	Key	Tokens in / out	Cost
14:02:31	claude-sonnet-5	production	18,420 / 1,206	$0.05379
14:02:07	claude-haiku-4-5	ci-review	6,310 / 412	$0.009208
14:01:44	gpt-6.1-sol	production	22,904 / 2,031	$0.07273
13:58:12	deepseek-v4-flash	batch	3,120 / 880	$0.001594

Preview of the console at oneaiapikey.com/console. Balances and requests are example data; costs use the published rates.

Keys: Create, name, limit and disable keys. A disabled key fails with 401 on its next call.
Usage: One row per request with model, key, tokens and cost.
Billing: Your balance in dollars and the credit packs you can add to it.

Open the console

Resources

Latest notes, news and guides.

Model releases that affect your code, with sources, and practical write-ups from running the gateway.

All resources

Oct 9, 2026Tutorials
Get Warned Before Your LLM Balance Runs Out: Alert Levels, Email and Signed Webhooks
Pick an alert level from your real daily spend, choose email or a webhook, and run a small signed receiver in Node or Python that you can test with curl.
Oct 8, 2026Product Updates
Agent Kits and Free Tools Are Live
Step-by-step setups for eight coding agents at /agents, and three free browser tools at /tools: a cost calculator, a config generator and a token estimator.
Oct 8, 2026Tutorials
Use the Anthropic Messages Format Through a Compatible Gateway
Call Claude through a gateway with the Anthropic SDKs: base URL, headers, streaming, a tool-use loop and prompt caching, in Python, Node and curl.

FAQ

Before you sign up.

More answers on the FAQ page, or write to support@relaykit.dev.

How do credits and billing work?

Pay as you go is prepaid: one credit equals $1 of usage at the published per-model rates. Each request deducts its cost based on the input, output and cached tokens it used, and every charge appears in your usage log. Pay as you go has no monthly fee and nothing renews; each top-up is $3 or more.

Are there monthly plans?

Plans are opening soon: Pro ($5 a month), Plus ($9 a month) and Max ($29 a month), each with a weekly allowance measured in DeepSeek V4 Flash input tokens. The allowance pays for any model at the plan rates (Plus 2% and Max 4% off every model). Until they open, every account pays as it goes from a prepaid wallet.

Are any models free?

Not right now. When a model is free or discounted it is labelled on the models pages, in Chat and on the pricing page, with the end date of the promotion.

Is there a free trial?

No. There is no sign-up credit. The smallest payment is $3: a first top-up of pay as you go, which works on every model, or, once plans open, a first month of Pro for $3, paid once by card and limited to one per account. Free models also become available once the account has paid $3 or more.

Which apps and coding tools can I use?

Anything that lets you set a custom OpenAI-compatible or Anthropic-compatible base URL. We publish step-by-step setup guides for Claude Code, Codex CLI, Cursor, Cline, Kilo Code, OpenCode, Aider and Zed, and the official OpenAI and Anthropic SDKs work by changing the base URL and key.

Is Relaykit affiliated with Anthropic, OpenAI or xAI?

No. Relaykit is an independent API platform. We are not affiliated with, endorsed by or sponsored by Anthropic, OpenAI, xAI or any other model provider. Model names are trademarks of their respective owners.

Create a key, change the base URL, send the request. Start from a $3 top-up.

Get API key Read the docs

One API key for Claude, GPT, Grok and DeepSeek.

Works with

What happens between your code and the model.

Your app

Key check

Routing

Provider

Metered response

Usage log