Docs
Kredo speaks the OpenAI Chat Completions and Responses APIs and the Anthropic Messages API. If a tool lets you set a base URL and an API key, it can use Kredo.
Base URL and keys
Every request goes to this base URL:
https://kredo.firebirds.cc/v1Create a key in the console under API keys. Send it as Authorization: Bearer <key>; the x-api-key header, which Anthropic-style clients use, works too. The examples below read the key from an environment variable named KREDO_API_KEY.
| Endpoint | What it is |
|---|---|
GET /v1/models | Models you can use, with limits and prices |
POST /v1/chat/completions | OpenAI Chat Completions |
POST /v1/responses | OpenAI Responses (used by Codex) |
POST /v1/messages | Anthropic Messages (used by Claude Code) |
All three POST endpoints work with and without streaming.
First request with curl
curl https://kredo.firebirds.cc/v1/chat/completions \
-H "Authorization: Bearer $KREDO_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "qwen3.8-max",
"messages": [{"role": "user", "content": "Say hello in one sentence."}],
"max_tokens": 200
}'To see which models your key can use:
curl https://kredo.firebirds.cc/v1/models \
-H "Authorization: Bearer $KREDO_API_KEY"OpenAI SDK
Use the official OpenAI libraries and change two things: the base URL and the API key. Everything else, streaming and tool calls included, works as you are used to.
import os
from openai import OpenAI
client = OpenAI(
base_url="https://kredo.firebirds.cc/v1",
api_key=os.environ["KREDO_API_KEY"],
)
stream = client.chat.completions.create(
model="qwen3.8-max",
messages=[{"role": "user", "content": "Write a haiku about caching."}],
max_tokens=200,
stream=True,
)
for chunk in stream:
if chunk.choices and chunk.choices[0].delta.content:
print(chunk.choices[0].delta.content, end="")import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://kredo.firebirds.cc/v1",
apiKey: process.env.KREDO_API_KEY,
});
const response = await client.responses.create({
model: "qwen3.8-max",
input: "Write a haiku about caching.",
max_output_tokens: 200,
});
console.log(response.output_text);Codex
Add a provider to ~/.codex/config.toml and make it the default:
model = "qwen3.8-max"
model_provider = "kredo"
[model_providers.kredo]
name = "Kredo"
base_url = "https://kredo.firebirds.cc/v1"
env_key = "KREDO_API_KEY"
wire_api = "responses"Then put your key in the environment and start Codex:
export KREDO_API_KEY="your key"
codexClaude Code
Claude Code talks to the Messages API. Point it at Kredo with environment variables. Note that the base URL here has no /v1 at the end; Claude Code adds it.
export ANTHROPIC_BASE_URL="https://kredo.firebirds.cc"
export ANTHROPIC_AUTH_TOKEN="your key"
export ANTHROPIC_MODEL="qwen3.8-max"
export ANTHROPIC_DEFAULT_OPUS_MODEL="qwen3.8-max"
export ANTHROPIC_DEFAULT_SONNET_MODEL="qwen3.8-max"
export ANTHROPIC_DEFAULT_HAIKU_MODEL="qwen3.8-flash"
export CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1
claudeThe three ANTHROPIC_DEFAULT_…_MODEL variables matter: Claude Code uses a small model for background work such as naming sessions, and without them it would ask for a model that does not exist here.
The same endpoint with curl:
curl https://kredo.firebirds.cc/v1/messages \
-H "x-api-key: $KREDO_API_KEY" \
-H "anthropic-version: 2023-06-01" \
-H "Content-Type: application/json" \
-d '{
"model": "qwen3.8-max",
"max_tokens": 200,
"messages": [{"role": "user", "content": "Say hello in one sentence."}]
}'Models
Use the model ID exactly as shown. Prices are on the pricing table.
| Model ID | Name | Max input tokens | Max output tokens | Images |
|---|---|---|---|---|
deepseek-v4-pro | DeepSeek V4 Pro | 128,000 | 65,536 | No |
glm-5.3 | GLM-5.3 | 128,000 | 65,536 | No |
qwen3-vl-plus | Qwen3-VL Plus | 128,000 | 65,536 | Yes |
qwen3.8-flash | Qwen3.8 Flash | 128,000 | 65,536 | No |
qwen3.8-max | Qwen3.8 Max | 128,000 | 65,536 | Yes |
qwen3.8-max-raw | Qwen3.8 Max Raw | 128,000 | 65,536 | Yes |
Billing and errors
Your balance is prepaid. Before each request, the most it could cost is set aside: the size of what you sent plus the largest reply the model may give. When the reply finishes, the real cost is charged and the rest is freed at once.
Set an output limit (max_tokens, max_completion_tokens or max_output_tokens) to keep the amount set aside small. Without one, the model's maximum output is assumed.
| Status | Meaning | Charged? |
|---|---|---|
400 | The request was not valid. The message says why. | No |
401 | The API key is missing, wrong or revoked. | No |
402 | Not enough credit to set aside the cost of this request. | No |
404 | The model ID does not exist. | No |
413 | The request body is too large. | No |
429 | Too many requests. Wait for the time in the Retry-After header. | No |
502 | The model provider failed or could not be reached. | No |
Errors use the format of the API you called: OpenAI-style on the OpenAI endpoints, Anthropic-style on /v1/messages. Every response carries an x-request-id header; quote it if you contact support.
If a streamed reply is cut off before it finishes, we cannot tell what it cost. The amount set aside for that request stays on hold until we have checked, and shows as "under review" in your usage. Questions: support@example.com.