Developer manual

Use DeepSeek through one DeepRelay key.

DeepRelay gives you an OpenAI-compatible endpoint and an Anthropic Messages-compatible endpoint. Create a sk-dr... key in Dashboard, add account credit, then point your SDK or coding tool at DeepRelay.

Create API key Set up coding tools llms.txt

Quick start

1. Get a key

Open Dashboard and create an API key.
Save the full key when it appears. It is shown once.
Keep the key outside source code. Use environment variables.

export DEEPRELAY_API_KEY="sk-drYOUR_KEY"

2. Call the API

curl https://deeprelayapi.cloud/v1/chat/completions \
  -H "Authorization: Bearer $DEEPRELAY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4-pro",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

A successful response means your key, balance, model route, and upstream channel are all working.

Endpoint summary

Interface	Base URL	Auth	Main endpoint
OpenAI compatible	`https://deeprelayapi.cloud/v1`	`Authorization: Bearer sk-dr...`	`POST /chat/completions`
Anthropic compatible	`https://deeprelayapi.cloud`	`x-api-key: sk-dr...` or tool token env	`POST /v1/messages`

Models

Model	Use it for	Recommended interface
`deepseek-v4-pro`	Coding, refactors, reasoning-heavy work, agent tasks.	OpenAI or Anthropic
`deepseek-v4-flash`	Fast answers, low-cost chat, simple coding tasks.	OpenAI or Anthropic
`deepseek-chat`	Legacy compatibility.	OpenAI
`deepseek-reasoner`	Legacy reasoning compatibility.	OpenAI

For coding tools, start with deepseek-v4-pro. Switch to deepseek-v4-flash only when speed and cost matter more than reasoning depth.

OpenAI-compatible API

Use this path when your SDK or framework already supports OpenAI Chat Completions.

Environment

export OPENAI_BASE_URL="https://deeprelayapi.cloud/v1"
export OPENAI_API_KEY="$DEEPRELAY_API_KEY"

Python

from openai import OpenAI

client = OpenAI(
    base_url="https://deeprelayapi.cloud/v1",
    api_key="sk-drYOUR_KEY",
)

response = client.chat.completions.create(
    model="deepseek-v4-pro",
    messages=[
        {"role": "user", "content": "Write a Python merge sort."}
    ],
)

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

Node.js

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://deeprelayapi.cloud/v1",
  apiKey: process.env.DEEPRELAY_API_KEY,
});

const response = await client.chat.completions.create({
  model: "deepseek-v4-pro",
  messages: [{ role: "user", content: "Explain rate limits." }],
});

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

Do not add another /v1 in SDK code when your base URL already ends with /v1.

Anthropic-compatible Messages API

Use this path for Claude Code and tools that expect Anthropic Messages. The base URL is https://deeprelayapi.cloud; the tool adds /v1/messages.

curl

curl https://deeprelayapi.cloud/v1/messages \
  -H "x-api-key: $DEEPRELAY_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{
    "model": "deepseek-v4-pro",
    "max_tokens": 1024,
    "messages": [
      {"role": "user", "content": "Hello from Messages API."}
    ]
  }'

Python Anthropic SDK

import anthropic

client = anthropic.Anthropic(
    base_url="https://deeprelayapi.cloud",
    api_key="sk-drYOUR_KEY",
)

message = client.messages.create(
    model="deepseek-v4-pro",
    max_tokens=1024,
    messages=[
        {"role": "user", "content": "Write a short TypeScript example."}
    ],
)

print(message.content[0].text)

If an Anthropic tool refuses custom model names, use the OpenAI-compatible path or set the tool's model override to deepseek-v4-pro.

One-click setup for AI coding tools

There are two setup paths. The AI-assisted path gives an agent exact instructions. The command-line path writes local config files for Claude Code and OpenCode.

AI-assisted setup

Copy this into Claude Code, OpenCode, Cursor Agent, Cline, or another local agent. Replace the key before sending it.

Prompt

Use https://deeprelayapi.cloud/docs/install/deeprelay.md to connect DeepRelay to my local AI coding tools.

My DeepRelay API key is: sk-drYOUR_KEY

Do not print the full key. Configure Claude Code and OpenCode if they are installed. Back up any existing config before changing it. After setup, verify with a low-cost model or a non-billing status command.

Command-line setup

Run this on your own machine. The script asks for your key and does not send it anywhere except DeepRelay validation calls.

Installer

curl -fsSL https://deeprelayapi.cloud/install/deeprelay.sh | bash

This is a local configuration helper, not account creation. Create your DeepRelay account and API key in Dashboard first.

Claude Code

Claude Code uses the Anthropic-compatible route. Clear old Anthropic variables first, then point Claude Code at DeepRelay.

Shell setup

unset ANTHROPIC_AUTH_TOKEN
unset ANTHROPIC_API_KEY
unset ANTHROPIC_BASE_URL

export ANTHROPIC_BASE_URL="https://deeprelayapi.cloud"
export ANTHROPIC_AUTH_TOKEN="$DEEPRELAY_API_KEY"
export ANTHROPIC_MODEL="deepseek-v4-pro"
export ANTHROPIC_DEFAULT_SONNET_MODEL="deepseek-v4-pro"
export ANTHROPIC_DEFAULT_OPUS_MODEL="deepseek-v4-pro"
export ANTHROPIC_DEFAULT_HAIKU_MODEL="deepseek-v4-flash"
export API_TIMEOUT_MS="3000000"
export CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC="1"

claude

`~/.claude/settings.json`

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://deeprelayapi.cloud",
    "ANTHROPIC_AUTH_TOKEN": "sk-drYOUR_KEY",
    "API_TIMEOUT_MS": "3000000",
    "CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC": "1",
    "ANTHROPIC_MODEL": "deepseek-v4-pro",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "deepseek-v4-pro",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "deepseek-v4-pro",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "deepseek-v4-flash"
  }
}

Verify inside Claude Code with /status and /model. The base URL should be DeepRelay and the model should be deepseek-v4-pro.

OpenCode

OpenCode can use provider configuration files and environment variables. The safest first setup is project-local: keep DeepRelay config in the repo you are coding in, not in a global shared file.

Environment

export DEEPRELAY_API_KEY="sk-drYOUR_KEY"
export OPENAI_BASE_URL="https://deeprelayapi.cloud/v1"
export OPENAI_API_KEY="$DEEPRELAY_API_KEY"

`opencode.json`

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "deeprelay": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "DeepRelay",
      "options": {
        "baseURL": "https://deeprelayapi.cloud/v1",
        "apiKey": "{env:DEEPRELAY_API_KEY}"
      },
      "models": {
        "deepseek-v4-pro": {},
        "deepseek-v4-flash": {}
      }
    }
  },
  "model": "deeprelay/deepseek-v4-pro"
}

OpenCode config support changes across releases. If your version rejects custom provider config, keep the environment variables above and choose an OpenAI-compatible provider in the TUI.

Cursor

WhereCursor Settings -> Models -> OpenAI API Key / Override OpenAI Base URL.

Base URLhttps://deeprelayapi.cloud/v1

API Keysk-dr...

Modeldeepseek-v4-pro for coding, deepseek-v4-flash for cheaper quick edits.

Cursor's OpenAI base URL override can affect other OpenAI-backed features. If built-in models stop working, remove the override and use DeepRelay only in a separate profile or tool.

Cline

Install or update the Cline extension in VS Code.
Open Cline settings and select an OpenAI-compatible provider.
Set Base URL to https://deeprelayapi.cloud/v1.
Set API Key to your sk-dr... key.
Set model to deepseek-v4-pro.

OPENAI_BASE_URL=https://deeprelayapi.cloud/v1
OPENAI_API_KEY=sk-drYOUR_KEY
MODEL=deepseek-v4-pro

TRAE

Open TRAE settings and go to the model/provider page.
Add a custom OpenAI-compatible provider.
Use https://deeprelayapi.cloud/v1 as the base URL.
Use your sk-dr... key and add deepseek-v4-pro as the model.
Run TRAE's built-in model test. If it fails, first verify the key with the curl command in Quick start.

Other tools

Tool type	Use this setting
OpenAI-compatible tools	Base URL `https://deeprelayapi.cloud/v1`, API key `sk-dr...`, model `deepseek-v4-pro`.
Anthropic-compatible tools	Base URL `https://deeprelayapi.cloud`, token/key `sk-dr...`, model `deepseek-v4-pro`.
Tools with fixed model menus	Choose custom model if available. If not, the tool may not support DeepRelay yet.

Streaming

curl https://deeprelayapi.cloud/v1/chat/completions \
  -H "Authorization: Bearer $DEEPRELAY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4-flash",
    "stream": true,
    "messages": [{"role": "user", "content": "Stream a short answer."}]
  }'

Common errors

Error	Likely cause	Fix
401	Wrong key, copied masked key, or deleted key.	Create a new key in Dashboard and save the full `sk-dr...` value.
Insufficient balance	Your DeepRelay account credit is empty.	Recharge before calling paid models.
Model not found	Wrong model name or a tool is forcing a native Claude/OpenAI model.	Use `deepseek-v4-pro` or `deepseek-v4-flash`.
Claude Code fails	Base URL includes extra `/v1` or old Anthropic env vars override the setup.	Set `ANTHROPIC_BASE_URL=https://deeprelayapi.cloud` and clear old vars.
Tool works but costs look confusing	Tools show tokens; DeepRelay bills by quota translated into account credit.	Use Dashboard Billing & Usage for money view and model-level usage.