From API key to first answer

1. Sign in to Redrob Console. Your first sign-in creates your workspace.
2. Open API keys and create a key.
3. Copy it there and then. The full key is shown once and never again.
4. Keep it server-side as REDROB_API_KEY.

The completion API speaks OpenAI’s chat completions protocol, so the official SDKs work unchanged. There is no Redrob package to install: set the base URL and the key.

# pip install openai
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["REDROB_API_KEY"],
    base_url="https://console.redrob.ai/api/backend/v1",
)

answer = client.chat.completions.create(
    model="claude-opus-5-5",
    messages=[{"role": "user", "content": "What is a robot?"}],
)

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

Anything built on those SDKs works the same way, such as LangChain, LlamaIndex and the Vercel AI SDK, wherever you can set an OpenAI-compatible base URL.

Worth doing once to confirm the key before any application code is involved.

export REDROB_API_KEY="rrk_…"

curl https://console.redrob.ai/api/backend/v1/chat/completions \
  -H "Authorization: Bearer $REDROB_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-opus-5-5",
    "messages": [{ "role": "user", "content": "What is a robot?" }]
  }'

A completion needs nothing special for this: ask in Tamil and the answer comes back in Tamil. Translating is a separate call, so you decide what gets translated and pay for exactly that, rather than every non-English request quietly costing two.

import httpx

# The completion itself needs nothing special: ask in Tamil, get Tamil back.
answer = client.chat.completions.create(
    model="claude-opus-5-5",
    messages=[{"role": "user", "content": "தமிழில் சென்னையைப் பற்றி சொல்லுங்கள்."}],
)

# Translate when you want to, and pay for just that.
english = httpx.post(
    "https://console.redrob.ai/api/backend/v1/translate",
    headers={"Authorization": f"Bearer {REDROB_API_KEY}"},
    json={
        "input": answer.choices[0].message.content,
        "sourceLanguage": "auto",
        "targetLanguage": "en",
    },
).json()

print(english["text"])

Fifteen languages: en, ko, hi, bn, pa, gu, or, ta, te, kn, ml, ur, mr, ne, as. Pass auto as the source and the language is detected by script, which costs nothing.

stream = client.chat.completions.create(
    model="claude-opus-5-5",
    messages=[{"role": "user", "content": "Explain the Indian monsoon."}],
    stream=True,
)

for chunk in stream:
    piece = chunk.choices[0].delta.content
    if piece:
        print(piece, end="", flush=True)

One caveat worth knowing: an answer that has to be translated back into the language it was asked in arrives as a single chunk, because there is nothing to show until the whole answer exists.

The same base URL also serves Anthropic’s Messages API at /v1/messages, so the Anthropic SDKs and Claude Code work too. Send the key as x-api-key or as a bearer token.

export ANTHROPIC_BASE_URL="https://console.redrob.ai/api/backend"
export ANTHROPIC_API_KEY="rrk_…"
export ANTHROPIC_MODEL="claude-sonnet-5-5"
export ANTHROPIC_SMALL_FAST_MODEL="claude-sonnet-5-5"

claude

Thinking is not streamed back, and tool calls arrive whole at the end of the turn.

• Every field and status code is in the API reference.
• Try prompts without writing code in the playground.
• Issue a key per environment in API keys, so one can be revoked without taking the others down.
• Watch tokens, cost, and latency per call in request logs.