Your API.
Your tools.

One key connects your applications to the ZAPRO.SU API.

1. Sign up and top up your balance in your account. You can start with a $5 gift from @neuroshop_markbot.
2. Open API Keys, click Create API Key, enter a name and select the AUTO group.
3. Save and copy the key. Then choose the guide for your application.

Choose an application

An AI assistant for code, files and images. Choose the setup method that suits you.

irm https://zapro.su/codex.ps1 | iex

curl -fsSL https://zapro.su/codex.sh | sh

Manual installation

The same API connection, configured through two files: config.toml and auth.json. You need an API key in the AUTO group.

1. Install Codex. On Windows/macOS, complete the initial sign-in if prompted, then fully quit the application, including its system tray icon.

Install Codex CLI on Linux

Run this in your system terminal:

TerminalCopy

curl -fsSL https://chatgpt.com/codex/install.sh | sh

Then configure the files as described below.

2. Open %USERPROFILE%\.codex on Windows or ~/.codex on macOS/Linux. Back up any existing config.toml and auth.json files.
3. Open or create config.toml and add these settings:

config.tomlCopy

model_provider = "zaprosu"
model = "gpt-6-astra"
model_reasoning_effort = "high"
cli_auth_credentials_store = "file"

[model_providers.zaprosu]
name = "ZAPROSU"
base_url = "https://po.zapro.su/v1"
wire_api = "responses"
requires_openai_auth = true
supports_websockets = false

If config.toml is already configured

Keep your other settings. Place the first four lines before the first [...] section, updating any existing values. Add [model_providers.zaprosu] once, or update the existing section. In that section, remove any old env_key, env_key_instructions, experimental_bearer_token, nested auth settings and Authorization header overrides: the key now comes from auth.json. If forced_login_method = "chatgpt" is set, remove that restriction to allow API-key authentication.

4. Create auth.json or replace its contents, inserting your own key:

JSONCopy

{
  "auth_mode": "apikey",
  "OPENAI_API_KEY": "sk-YOUR-KEY"
}

5. Save the files and start the application; switch Work / ChatGPT to Codex if needed. On Linux, run codex. Create a new chat and send a short prompt.

Image generation#

Install the Provider Image skill. The automatic installer already includes it.

Windows — PowerShell:

Windows · PowerShellCopy

irm https://zapro.su/pi | iex

macOS / Linux — Terminal:

TerminalCopy

curl -fsSL https://zapro.su/pi.sh | sh

Fully restart Codex and type: “Generate an image of a cat wearing boots.” To edit an image, attach it and describe the changes. The skill is available under Plugins / Manage.

Restore your chats

If old chats disappear after switching providers, ask Codex to restore their visibility. The automatic installer handles migration for you.

TEXTCopy

Restore the visibility of my chats after switching from the openai provider to zaprosu.
First, find the current database and session files in my .codex folder and inspect their structure.
Make a consistent backup of the database and all files to be modified; stop processes using them before writing.
Change only the model_provider field from openai to zaprosu in the relevant chats and sessions.
Do not touch other providers, keys or settings.
Check the integrity of the database and modified JSON/JSONL files. If a safe write is not possible, stop and explain why.

If you cannot sign in or connect, contact support.

1. Download and install Chatbox: Windows / macOS.
2. Open Settings, then “Model Provider”, then “Claude”.
3. Enter your API key.
4. Set API Host to https://po.zapro.su/v1.
5. Click “Check” and select Claude Opus 4.6.
6. Next to Model, click “Fetch” and select Claude Opus 4.8 from the list.
7. Return to the application's main screen.
8. Start a new chat.
9. Select Claude Opus 4.8 from the model list.
10. Start chatting with the model on your computer.

Image generation in Chatbox#

1. Open Settings → Model Provider → OpenAI.
2. Enter your API key and set API Host to https://po.zapro.su/v1.
3. Click Fetch, add gpt-image-2, open its settings using the gear icon and select the Image model type.

4. Close settings and click Create Image.

5. Select the OpenAI provider and the GPT Image 2 model.
6. Enter your prompt.

You can also use this approach with the OpenAI provider and other models available through this service.

1. Enable a VPN. We recommend this Telegram bot (one week of free access).
2. Download Claude Desktop, install it and launch the application.
3. On the sign-in screen, open the top-left menu: Help → Troubleshooting → Enable Developer Mode.
4. In the new Developer menu, select Configure Third-Party Inference. Under Connection, enter:

5. Click Test connection: a green result means the connection works. Below, click Test model discovery — models will appear under Model discovery — found…. You do not need to add them through Add model.
6. On the left, open Workspace and click Chat to add it.
7. At the bottom, click Apply Changes → Save & Restart.

After restarting, open a chat and select a model.

Create an API key#

First, create a new key in the dashboard. You can select the Claude group, but AUTO is recommended for compatibility with more models. Use this key in the steps below.

Install Claude Code if needed#

Windows PowerShell (press Win + ~)

irm https://claude.ai/install.ps1 | iex

macOS (open the Terminal application)

curl -fsSL https://claude.ai/install.sh | bash

Edit the configuration#

Open the Claude folder: on Windows, press Win + R, enter %USERPROFILE%\.claude and press Enter; on macOS, press Cmd + Shift + G, enter ~/.claude and press Enter.

Find settings.json, open it in a text editor and replace its contents with:

{
  "autoUpdatesChannel": "stable",
  "env": {
    "ANTHROPIC_AUTH_TOKEN": "sk-YOUR-CLAUDE-KEY",
    "ANTHROPIC_BASE_URL": "https://po.zapro.su",
    "ANTHROPIC_DEFAULT_FABLE_MODEL": "claude-fable-5",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "claude-haiku-4-5-20251001",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "claude-opus-5",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "claude-sonnet-5",
    "CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC": "1",
    "CLAUDE_CODE_ENABLE_TELEMETRY": "0",
    "DISABLE_AUTOUPDATER": "1",
    "DISABLE_COST_WARNINGS": "1",
    "DISABLE_ERROR_REPORTING": "1",
    "DISABLE_TELEMETRY": "1"
  },
  "model": "claude-opus-5",
  "permissions": {
    "allow": [],
    "defaultMode": "bypassPermissions",
    "deny": []
  },
  "skipDangerousModePermissionPrompt": true
}

bypassPermissions and skipDangerousModePermissionPrompt are optional. They disable permission prompts; you can remove these lines.

Then open a terminal in any folder and run:

claude

An AI assistant in your editor. Choose the Claude or Codex extension.

Claude

Claude in the editor sidebar. The extension does not require a separate CLI installation. Have your API key in the AUTO group ready.

1. In VS Code, open Extensions and install Claude Code by Anthropic.
2. Open the command palette: Ctrl + Shift + P (Windows/Linux) or Cmd + Shift + P (macOS). Select Preferences: Open User Settings (JSON).
3. Add these settings, replacing sk-YOUR-CLAUDE-KEY with your key:

JSONCopy

{
  "claudeCode.disableLoginPrompt": true,
  "claudeCode.environmentVariables": [
    { "name": "ANTHROPIC_AUTH_TOKEN", "value": "sk-YOUR-CLAUDE-KEY" },
    { "name": "ANTHROPIC_BASE_URL", "value": "https://po.zapro.su" },
    { "name": "CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY", "value": "1" },
    { "name": "ANTHROPIC_DEFAULT_FABLE_MODEL", "value": "claude-fable-5" },
    { "name": "ANTHROPIC_DEFAULT_HAIKU_MODEL", "value": "claude-haiku-4-5-20251001" },
    { "name": "ANTHROPIC_DEFAULT_OPUS_MODEL", "value": "claude-opus-5" },
    { "name": "ANTHROPIC_DEFAULT_SONNET_MODEL", "value": "claude-sonnet-5" },
    { "name": "CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC", "value": "1" },
    { "name": "CLAUDE_CODE_ENABLE_TELEMETRY", "value": "0" },
    { "name": "DISABLE_AUTOUPDATER", "value": "1" },
    { "name": "DISABLE_COST_WARNINGS", "value": "1" },
    { "name": "DISABLE_ERROR_REPORTING", "value": "1" },
    { "name": "DISABLE_TELEMETRY", "value": "1" }
  ]
}

If the JSON file already has content, keep your other settings: add only the two claudeCode.* properties inside the existing { }, with the required commas. Update existing properties without creating duplicates.

4. Save the file, fully restart VS Code, open your project folder and the Claude Code panel. Enter /model and select a model.

Choosing a model

The model list loads automatically for your key. You can specify a full name: /model claude-opus-5. The aliases fable, opus, sonnet and haiku use the mappings above. If a model is missing, check that your key uses the AUTO group and restart VS Code. Model availability depends on your key and the service.

An AI coding agent. Connect OpenCode Desktop through its interface — no terminal or file editing needed. You need an API key in the AUTO group and a positive balance.

1. Open the OpenCode download page. Under OpenCode Desktop, download the application for Windows, macOS or Linux, install it and launch it.
2. Open Settings → Providers. Next to Custom provider, click Connect.
3. Fill in the connection fields and one row under Models:

4. Leave Headers (optional) empty and click Submit. The application saves the settings for you.
5. Use Open project to choose your project folder, open a new chat and select GPT-6 Astra from ZAPRO.SU in the model list. Type “Reply OK” or describe your task.

To add other models, use Add model: enter the exact name from your API model catalog in ID and a display label in Name. This connection uses Chat Completions; do not append /chat/completions to the Base URL.

OpenCode documentation: custom providers.

A terminal agent for files, code and tasks. You need an API key in the AUTO group and a positive balance.

1. Install Hermes using the command for your system.

Windows — PowerShell:

iex (irm https://hermes-agent.nousresearch.com/install.ps1)

macOS / Linux — Terminal:

curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash

On macOS/Linux, Git must already be installed; Linux also requires curl and xz-utils. The installer sets up the remaining dependencies.

2. After installation, open a new terminal window and start model setup:

hermes model

3. Select Custom endpoint and fill in the fields:

• API mode / transport: Chat Completions (chat_completions).
• API base URL: https://po.zapro.su/v1.
• API key: your ZAPRO.SU key.
• Model name: gpt-6-astra.
• If asked for a connection name: ZAPROSU.

4. Save the settings and start a chat:

hermes

Type “Reply OK”. To change the connection, exit the chat and run hermes model again. If the setup wizard opens during installation, enter the same settings there.

Official Hermes installation guide · Custom API setup.

A personal AI agent with a browser-based chat. You need an API key in the AUTO group and a positive balance.

1. Install OpenClaw using the command for your system.

Windows — PowerShell:

iwr -useb https://openclaw.ai/install.ps1 | iex

macOS / Linux — Terminal:

curl -fsSL https://openclaw.ai/install.sh | bash

2. In the wizard, select Custom setup, then Custom Provider (it may be under More…). If the wizard does not start, open a new terminal and run:

openclaw onboard

3. Enter the connection settings:

• Endpoint compatibility: OpenAI-compatible (/chat/completions).
• Base URL: https://po.zapro.su/v1.
• API key: your ZAPRO.SU key.
• Model ID: gpt-6-astra.
• Optional name / alias: ZAPROSU.

Wait for a successful model check and finish local setup (Local, if the wizard offers a choice). You can skip messengers and additional plugins for your first chat.

4. In the browser dashboard that opens, type “Reply OK”. If it does not open, run this in another terminal:

openclaw dashboard

Keep the Gateway terminal open while using the agent.

openclaw gateway install
openclaw gateway status
openclaw dashboard

Official OpenClaw installation guide · Custom API setup.

Installation#

Download for Android / iPhone.

1. Open Settings, then “Model Provider”, then “OpenAI” (the standard option, not Responses) or “Claude”.
2. Enter your API key.
3. Set API Host to https://po.zapro.su/v1.
4. Click “Check” and select GPT-5.4.
5. Next to Model, click “Fetch” and select the models you want to add, such as GPT-5.5 or claude-opus-5.
6. Return to the main screen.
7. Start a new chat.
8. Select the model you want from the list.
9. Start using it on your phone.

For any questions, contact https://t.me/aipro_helper