Quickstart

Call Corion AI in a few minutes. One base URL, one key, and whichever API your client already speaks: OpenAI or Anthropic.

Corion AI is in MVP preview. Model availability and commercial terms remain subject to launch configuration.

1. Create an account

Sign up, then confirm your email if your workspace requires confirmation.

2. Add credits

Every new account starts with a welcome credit. Open Dashboard → Billing to check your balance or buy an available credit pack. Credit packs fund usage; they are not subscriptions. Test deployments use Stripe test mode and do not create a real charge. When your balance runs out, the playground prompts you to add credits.

3. Create an API key

Open Dashboard → API keys, name the key, and choose Create key. Copy the complete key immediately. Corion stores only the key identifier and a display prefix, so the complete secret cannot be shown again.

Keep keys in a server-side secret store. Never embed one in browser code or commit it to source control.

4. Pick your API surface

Everything below lives under https://api.corion.ai, takes the same key, and bills against the same credits. Choose the one your client already speaks. No adapter or shim sits in between.

Both Authorization: Bearer <key> and Anthropic's x-api-key: <key> are accepted on every surface, so SDK defaults work unchanged.

Available model identifiers are kimi-k3, deepseek-v4-pro-0813 and deepseek-v4-flash-0731. See pricing for the per-model rates charged against your credits.

5. Make the first request

export CORION_BASE_URL="https://api.corion.ai"
export CORION_API_KEY="your-key"

curl "$CORION_BASE_URL/v1/chat/completions" \
  -H "Authorization: Bearer $CORION_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kimi-k3",
    "messages": [{"role": "user", "content": "Write a two-line launch announcement."}]
  }'

The same key over Anthropic Messages. Note the anthropic-version header and max_tokens, which that API requires:

curl "$CORION_BASE_URL/v1/messages" \
  -H "x-api-key: $CORION_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kimi-k3",
    "max_tokens": 256,
    "messages": [{"role": "user", "content": "Write a two-line launch announcement."}]
  }'

A 200 with an Anthropic-shaped body (type: "message", content: [{type: "text", ...}]) means auth and routing both work.

6. Stream a response

Both official SDKs take a custom base URL, so pointing them at Corion is a one-line change.

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["CORION_API_KEY"],
    base_url=f'{os.environ["CORION_BASE_URL"]}/v1',
)

stream = client.chat.completions.create(
    model="kimi-k3",
    messages=[{"role": "user", "content": "Explain speculative decoding simply."}],
    stream=True,
)

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

The Anthropic SDK reaches the same models over /v1/messages. Pass the base URL without /v1, because the SDK appends it:

import os
import anthropic

client = anthropic.Anthropic(
    api_key=os.environ["CORION_API_KEY"],
    base_url=os.environ["CORION_BASE_URL"],
)

with client.messages.stream(
    model="kimi-k3",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Explain speculative decoding simply."}],
) as stream:
    for text in stream.text_stream:
        print(text, end="", flush=True)

System prompts, tool use, and streaming all behave as each provider documents them. Corion serves the same wire formats.

7. Connect a coding agent

Most agents need two lines and no adapter, because Corion serves the protocol each one expects:

# Claude Code: Anthropic Messages
export ANTHROPIC_BASE_URL=https://api.corion.ai
export ANTHROPIC_API_KEY=$CORION_API_KEY
export ANTHROPIC_MODEL=kimi-k3
claude

# Kimi Code: imports every Corion model from our registry
kimi provider add https://corion.ai/kimi-registry/api.json --api-key $CORION_API_KEY
kimi -m corion/kimi-k3

Codex needs one ~/.codex/config.toml entry instead, because it speaks only the Responses API to custom providers:

model_provider = "corion"
model = "kimi-k3"

[model_providers.corion]
base_url = "https://api.corion.ai/v1"
env_key = "CORION_API_KEY"

Full setup, model switching, and troubleshooting per agent: Claude Code, Codex, Kimi Code, Hermes, OpenClaw, OpenCode.

Whichever agent you use, run a read-only task first ("summarize this repository") and confirm both the reply and its tool calls before granting permission to edit files.

8. Try the playground

The playground uses your signed-in account and available credits. It does not expose your virtual API key to browser JavaScript.

Continue with the API reference, then review errors and retries before integrating a production workload.