FAQs

Integration

What is MitDotKey?

MitDotKey is an AI Gateway compatible with the OpenAI SDK and Anthropic Messages. Users can connect AI and coding tools through one shared endpoint to call supported models.

Which tools does MitDotKey support?

The documentation lists tool groups such as VS Code + Claude, Claude Code, Claude App, Codex/OpenAI CLI, Cline/Roo, Continue.dev, Cursor, OpenWebUI/Lobe, Python/Node SDK, and direct API usage.

What is the shared Base URL?

The shared Base URL for both OpenAI and Anthropic usage is: https://api.mitdotkey.com. 

See more: https://mitdotkey.com/docs

What is the OpenAI-style authentication header?

Use this header: Authorization: Bearer mdk-… Here, “mdk-…” is your API key.

What is the Anthropic-style authentication header?

Use this header: x-api-key: mdk-… Here, “mdk-…” is your API key.

Can I use the same Base URL for the OpenAI SDK and Anthropic Messages?

Yes. The documentation states that “https://api.mitdotkey.com” is shared by OpenAI and Anthropic usage; the main difference is the authentication header style.

Payment

When do I need to top up?

Models under “aws/*” and “enterprise/*” deduct from the USD balance. If the balance is 0, go to Dashboard → Topup to add funds.

Is there a free option?

You can use the “free/*” tier. The documentation states that the free tier provides 400 requests per day for users who have previously topped up.

Starter

How many steps are needed for the first run?

There are three main steps: get an API key, top up if there is no balance, then quickly test the key and base URL using a terminal or the Playground.

Where do I get an API key?

Go to Dashboard → API Keys, click + Create in the top-right corner, give the key a memorable name, and click Create.

What does a MitDotKey API key look like?

The new key appears in the API Keys list. According to the documentation, every key starts with the prefix “mdk-”.

What security precautions should I take after creating an API key?

Copy and store the key immediately, and treat it like a password. Do not commit it to git or paste it into public chats. If it is exposed, delete it and create a new one.

How do I test the key in a terminal?

Call GET /v1/models to check the model list, then call /v1/chat/completions for a chat smoke test. Replace “mdk-your-api-key” with your real key.

What does a successful test look like?

The list-models command should return HTTP 200 and a JSON model list. The chat smoke test should return a response. If both pass, the key and endpoint are working.

Where can I test if I do not have a terminal?

You can use Dashboard → Playground to try chatting directly on the web.

Common errors

What usually causes a 401 / Unauthorized error?

It is usually caused by a wrong key, a missing “Bearer ” prefix before the key in the Authorization header, or a deleted key.

What usually causes a 404 / model_not_found error?

It is usually caused by an incorrect model name. Copy the exact model ID from the API & Models tab; for Opus/Sonnet, the effort tier such as “-medium” may be required.

How do I handle a 402 / Insufficient balance error?

This error means the balance is depleted. Top up in Topup, or switch to a “free/*” model if appropriate.

What usually causes Connection refused / timeout?

It is often caused by a Base URL missing “/v1” or a mistyped domain. Check that the Base URL is exactly “https://mitdotkey.com/v1”.

AI Models

Which model is suitable for daily coding or chat?

The documentation recommends “aws/claude-sonnet-4-6-medium” for daily coding and chat.

Which model is suitable for hard tasks or deep reasoning?

The documentation recommends “aws/claude-opus-4-8-medium” for hard tasks, agents, or deep reasoning.

Which models are cheap and fast?

The documentation suggests cheap and fast options such as “aws/qwen3-codex” and “aws/claude-haiku-4-5”.

Which model can I use for free testing?

The documentation suggests “free/claude-haiku-4-5” for free testing.

Where can I view realtime pricing and the model list?

You can view them at Dashboard → Models or call GET /v1/models. The documentation also notes that the full model list is in the API & Models tab.

Do I need to enter the model ID exactly?

Yes. An incorrect model name can trigger “model_not_found”. Copy the exact model ID from API & Models or from the GET /v1/models response.