Quickstart

The Anthropic Base URL does not include /v1; Anthropic clients append the API path. The OpenAI Base URL must include /v1.

Setup steps

Three steps: sign up, create a key, and put the Base URL and key into your tool or code.

Send your first request

First list the models the current key can call:

curl https://api.hop-base.com/v1/models \
  -H "Authorization: Bearer sk-your-key"

Then send a minimal chat request. The curl, Python, Node.js, Go, and Java tabs call gpt-5.6-sol over the OpenAI protocol and need a Codex group key; the Claude tab calls claude-sonnet-5 over the Anthropic protocol and needs a Claude group key. For another model family, replace model with an ID that /v1/models returns for that key.

curl https://api.hop-base.com/v1/chat/completions \
  -H "Authorization: Bearer sk-your-key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.6-sol",
    "messages": [{"role": "user", "content": "Reply with only: ok"}]
  }'

A successful call returns HTTP 200. The OpenAI protocol puts the reply in choices[0].message.content, the Anthropic protocol in content[0].text, and both include a usage object:

{
  "object": "chat.completion",
  "model": "gpt-5.6-sol",
  "choices": [{
    "index": 0,
    "message": { "role": "assistant", "content": "ok" },
    "finish_reason": "stop"
  }],
  "usage": { "prompt_tokens": 12, "completion_tokens": 2, "total_tokens": 14 }
}

A 401 means the key is wrong. A 404 saying the group does not support the model means none of the key's groups serves that model — use an ID from /v1/models, or add the matching group to the key.

To have a coding agent make the project changes, continue to Integrate with AI.

Balance and billing

• Prepaid: top up your account under Recharge in the console sidebar; each call is deducted from the balance by actual usage. Once the balance runs out, new requests return 402.
• Rates: the signed-in model catalog is authoritative; every charge is itemised under Usage in the console.
• Failures are not charged: failed requests are not billed in principle; a stream that breaks midway is billed for the output already produced. Full rules: Concurrency, timeouts, and billing.

Which group to choose

Choose groups by the models you want to call. One key can include 1–10 groups, and each request is billed at the rate of the group that serves its model. The key dialog lists only currently sold groups and does not let you combine groups whose models overlap (except GPT Official Direct (Azure) with a Codex group or GPT Image (all models) — see the note after the GPT table); a subscription group must be a key's only group. Rates depend on your account and are authoritative in the signed-in model catalog — these tables only show which group serves which models. Group names below are as the console shows them in English.

Claude Code and the Claude API

Chat and coding (OpenAI protocol)

Mixing Azure with Codex and GPT Image: one key can include GPT Official Direct (Azure) together with either Codex Plus or Codex Pro, and can also add GPT Image (all models); you do not need to pick a source.

• Models in both groups go to the Azure group first, whatever order you selected the groups in. The models shared with the Codex groups are gpt-5.5, gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna, gpt-6-sol, gpt-6-luna, gpt-6-astra, gpt-6.1-sol; the ones shared with GPT Image (all models) are gpt-image-2.5-flare and gpt-image-2.5-sunburst. If Azure fails before returning any output, the request switches to the other group automatically. A response that has already started is not switched if it breaks off.
• Each request is billed by the group that actually served it, and usage records show that group.
• Models in only one group are served by that group as usual — for example, codex-* is only in the Codex groups, gpt-5.4 and gpt-4.1 are only in the Azure group, and gpt-image-2 is only in GPT Image (all models).
• Requests served by the Azure group do not support the Fast (priority) tier; they run and bill at the standard tier.
• The key dialog, key list, and connect dialog label the two groups Primary and Backup, with a one-line billing note.

Other groups whose models overlap, such as Codex Plus and Codex Pro, still cannot share a key.

Image generation

Gemini Banana images live in the two Gemini groups above, Grok images in Grok Suite, and Kling images in Kling Official below.

Video generation

Speech

Next steps

• OpenAI SDK, Anthropic SDK, curl — full chat, streaming, and tool-calling examples
• Models — model IDs and capabilities
• Concurrency, timeouts, and billing — concurrency caps, timeouts, and billing rules
• Error codes and retries — what each error code means and whether to retry