Openrouter ICU User Guide

Updated today

Related link: Openrouter ICU Claude Code User Guide

Related link: Openrouter ICU Gemini User Guide

Related link: Openrouter ICU Recharge Agent Tutorial

Related link: Gemini Pro Ready-Made Account User Guide

Supported models:

OpenRouter ICU API User Guide

1. Log in and get an API key

Step 1:

Go to https://tokenadvent.com and log in. Enter your account and password. If you do not have an account, register one with your email address.

Step 2:

Click Redeem in the left navigation bar, then enter your redemption code.

Step 3:

Click API Keys in the left navigation bar, then click Create Key on the right side of the page. You only need to set a name and a group, then click Create.

(The name can be anything. For the group: choose OpenAI Balance for usage cards, openai subscription for the new monthly card, and Subscription for the old monthly card.)

Step 4:

Click Use Key to view how this API key should be used and configured in different applications.

2. Configuration methods for different applications

Cursor third-party API setup: Cursor Third-Party API Configuration – Token Advent

Codex third-party API setup: Cursor Third-Party API Configuration – Token Advent

Claude Code third-party API setup: Claude Code third-party API setup

Configure a third-party API in the VS Code Codex extension: Configure a third-party API key in the VS Code Codex extension

Configure a third-party API in IntelliJ IDEA AI Assistant: Configure a third-party API in IntelliJ IDEA AI Assistant

OpenClaw third-party API setup: OpenClaw third-party API setup

Hermes third-party API setup: Hermes third-party API setup

ccswitch third-party API setup: ccswitch third-party API setup

TRAE third-party API setup: TRAE third-party API setup

CherryStudio third-party API setup: Cherry Studio third-party API setup

3. Tutorial video

(1) Codex installation tutorial

4. FAQ

(1) Can the API key subscription from our platform share conversation history with your official Codex subscription?

Yes. Follow the guide to configure config.toml, and the conversation history should be inherited automatically. If automatic inheritance fails, try the manual recovery options below.

---

• Option 1:

Open the Codex root directory, which is the folder containing config.toml. In the /sessions folder, find the conversation record you want to restore.

Conversation records are stored in this format: /sessions/year/month/day/rollout-xxx.xxx.jsonl, as shown below.

Find the model_provider field in the first line of the conversation record. Following Option 1, set config.toml to the same provider name as the conversation record, such as OpenAI. The name is case-sensitive. Then restart Codex.

If model_provider does not exist, use provider instead.

---

• Option 2:

Conversation history inheritance only depends on model_provider. If the previous method does not work, there are usually two possible reasons:

1. The operation was performed incorrectly.
2. The old and new providers are completely incompatible. As the final fallback, feed the conversation record file to a new Codex thread to recover as much context as possible. The file is in the .codex/sessions folder, for example ~/.codex/sessions/xxxxxxxx.

---

(2) When using Codex with an API key from our platform, how do I enable or disable fast mode?

Codex CLI users can follow the Codex CLI prompt to enable or disable fast mode. Codex App users currently cannot enable fast mode. You can set the reasoning mode to low to make it a bit faster.

(3) Common errors and self-check fixes

• 401 error:
  In about 80% of cases, this is a configuration-file issue. Check the contents of config.toml and auth.json in the .codex folder. If you previously logged in with a GPT account, the contents will definitely be different. You need to configure them strictly according to the format in Codex third-party API setup. Also check that the file extensions are actually .toml and .json.

In about 10% of cases, the cause is the network environment. Check whether your proxy is turned off and whether you can directly access our official website. If you have checked the above and still cannot connect, contact customer support.

The final 10% is occasional key failure. Create a new key and try again.

• 403 error:
  The common causes are: you bought a usage card but did not redeem the card number, selected the wrong group for a monthly card, or the balance quota has been used up. If you have checked and corrected these issues but still cannot connect, contact customer support.

Usage card not redeemed:

Wrong monthly-card group selected:

• 429 error:
  Usually your quota has been used up. Go to the official website to check your remaining balance.

• 502 error:
  Usually the context is too long and exceeds the official compression limit of 258k. Try opening a new window and running the task again.
• 524 error:
  Try opening a new window and starting a test conversation. If the error still appears, replace the URL in the configuration file.

https://tokenadvent.com
# In some software, you may need to add /v1

(4) opencode shows Bad Request while running

If opencode shows Bad Request: {"detail":"Instructions are required"}, use the following fix:

"npm":"@ai-sdk/openai-compatible"

Add this line to resolve the error.