DEVSAIO

API documentation

DEVSAIO is one HTTPS API for the AI models listed on the model directory. It speaks the OpenAI format (chat completions, Responses, images) and the Anthropic Messages format, so existing SDKs and tools work by changing the base URL and key.

Quickstart

  1. Create an account and an API key. Keys are created on API Keys and start with dsk_. The full key is shown once.
  2. Add balance. Usage is prepaid. Top up on Billing.
  3. Send a request to https://devsaio.lol/v1 with your key:
curl https://devsaio.lol/v1/chat/completions \
  -H "Authorization: Bearer $DEVSAIO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "grok-4.6",
    "messages": [{"role": "user", "content": "Say hello in one sentence."}]
  }'
# pip install openai
from openai import OpenAI

client = OpenAI(base_url="https://devsaio.lol/v1", api_key="dsk_...")

r = client.chat.completions.create(
    model="grok-4.6",
    messages=[{"role": "user", "content": "Say hello in one sentence."}],
)
print(r.choices[0].message.content)
// npm install openai
import OpenAI from "openai";

const client = new OpenAI({ baseURL: "https://devsaio.lol/v1", apiKey: process.env.DEVSAIO_API_KEY });

const r = await client.chat.completions.create({
  model: "grok-4.6",
  messages: [{ role: "user", content: "Say hello in one sentence." }],
});
console.log(r.choices[0].message.content);

Authentication

Every request needs an API key. Send it in either header:

HeaderFormatUsed by
AuthorizationBearer dsk_…OpenAI SDKs, cURL, most tools
x-api-keydsk_…Anthropic SDKs and Claude Code
Keep keys server-side. Never ship a key in a browser or mobile app bundle. DEVSAIO stores only a hash of each key; if you lose it, create a new one and revoke the old one.

Model identifiers

Use the exact id from GET /v1/models or the model directory in the model field. Ids are lowercase, for example grok-4.6. Only models an administrator has published and priced are callable; anything else returns 404 model_not_found.

GET List models

GEThttps://devsaio.lol/v1/models

Returns every model you can call right now, in the OpenAI list format.

curl https://devsaio.lol/v1/models \
  -H "Authorization: Bearer $DEVSAIO_API_KEY"
{
  "object": "list",
  "data": [
    { "id": "grok-4.6", "object": "model", "created": 0, "owned_by": "…" }
  ]
}

POST Chat completions

POSThttps://devsaio.lol/v1/chat/completions

OpenAI-compatible. The request body is forwarded to the model unchanged apart from the model id, so the parameters a model supports upstream (such as tools, response_format or temperature) work as documented by its provider.

FieldTypeRequiredNotes
modelstringYesA model id from /v1/models.
messagesarrayYesObjects with role and content.
streambooleanNoServer-sent events. See Streaming.
max_tokensintegerNoCaps the output. Also used to size the balance hold.

POST Anthropic Messages

POSThttps://devsaio.lol/v1/messages

The native Anthropic format, used by Claude Code and the Anthropic SDKs. The anthropic-version and anthropic-beta headers are passed through, errors come back in Anthropic's format, and POST /v1/messages/count_tokens is available free of charge.

# pip install anthropic
import anthropic

client = anthropic.Anthropic(base_url="https://devsaio.lol", api_key="dsk_...")
msg = client.messages.create(
    model="grok-4.6",
    max_tokens=512,
    messages=[{"role": "user", "content": "Hello"}],
)
print(msg.content[0].text)
export ANTHROPIC_BASE_URL="https://devsaio.lol"
export ANTHROPIC_AUTH_TOKEN="dsk_..."
claude

POST Responses (Codex)

POSThttps://devsaio.lol/v1/responses

The OpenAI Responses API, which Codex CLI uses. Configure Codex with wire_api = "responses"; the Codex guide generates the full config.toml.

POST Image generation

POSThttps://devsaio.lol/v1/images/generations

OpenAI images format. Image models are billed per image, at the price shown on the pricing page.

curl https://devsaio.lol/v1/images/generations \
  -H "Authorization: Bearer $DEVSAIO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "IMAGE_MODEL_ID", "prompt": "A lighthouse at dusk", "n": 1}'

POST Video generation

POSThttps://devsaio.lol/v1/videos

Videos are generated in the background. Create a job, poll it until status is completed, then download the MP4. Video models are billed per second of video (the seconds you request, default 5) when the job is created. If a job fails, the charge is refunded automatically the first time you poll it.

curl https://devsaio.lol/v1/videos \
  -H "Authorization: Bearer $DEVSAIO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "VIDEO_MODEL_ID", "prompt": "A paper boat sailing down a rainy street", "seconds": "5", "size": "1280x720"}'
curl https://devsaio.lol/v1/videos/VIDEO_ID \
  -H "Authorization: Bearer $DEVSAIO_API_KEY"
# → {"id": "...", "status": "in_progress", "progress": 40}
curl -L https://devsaio.lol/v1/videos/VIDEO_ID/content \
  -H "Authorization: Bearer $DEVSAIO_API_KEY" -o video.mp4

Prefer a UI? Use Video Studio in your dashboard.

Streaming

Set "stream": true to receive server-sent events as the model writes. Chunks are passed through as the upstream sends them, in OpenAI or Anthropic format depending on the endpoint. Usage is read from the final event, and you are charged once when the stream ends. If the stream fails before any usage is reported, you are not charged.

stream = client.chat.completions.create(
    model="grok-4.6",
    messages=[{"role": "user", "content": "Count to five"}],
    stream=True,
)
for chunk in stream:
    print(chunk.choices[0].delta.content or "", end="")

Errors

Errors use the OpenAI shape {"error": {"message", "type", "code"}}, or Anthropic's shape on /v1/messages. Requests that fail are not charged.

StatusCodeMeaning and fix
400invalid_json, missing_modelThe body is not JSON or has no model.
401invalid_api_keyMissing, wrong or revoked key, or the account is suspended.
402insufficient_fundsBalance too low for the request. Top up.
404model_not_found, endpoint_not_availableUnknown model id, or an endpoint not enabled on this server.
409idempotency_conflictAn Idempotency-Key was reused. The first request stands; you were not charged twice.
429rate_limit_exceeded, key_limit_exceededToo many requests per minute (see Retry-After), or the key's monthly spending limit was reached.
502upstream errorThe model provider failed. Retry later; nothing was charged.
503model_unpricedThe model is listed but has no price yet.

Idempotency and retries

Send a unique Idempotency-Key header (for example a UUID per logical request). If a network error makes you retry with the same key, DEVSAIO refuses the duplicate with 409 instead of running and billing it twice.

curl https://devsaio.lol/v1/chat/completions \
  -H "Authorization: Bearer $DEVSAIO_API_KEY" \
  -H "Idempotency-Key: order-1042" \
  -H "Content-Type: application/json" \
  -d '{"model": "grok-4.6", "messages": [{"role": "user", "content": "Hi"}]}'

Usage and billing

  • Prepaid. Calls are paid from your balance. There is no subscription.
  • Per-token or per-image prices are shown on the pricing page and in your dashboard, in USD per million tokens or per image.
  • Hold, then exact charge. Before a call, an estimated amount is held from your balance. When the call finishes, you are charged the exact cost from the reported token usage and the rest of the hold is released.
  • Prompt caching. When a provider reports cached input, cache reads are billed at 10% of the input price and cache writes at 125%, matching how the providers price them.
  • Failed requests are free. Errors and upstream failures release the hold without a charge.
  • Every call, with tokens, latency and cost, is in Usage Logs and can be exported to CSV.

API key management

On API Keys you can create keys per project, set an optional monthly spending limit on each, see each key's spend this month and when it was last used, and revoke a key immediately. A revoked key stops working on the next request.

Integration guides

Ready-made settings, with your key filled in when you are signed in:

What we store

  • API requests: billing metadata only (model, token counts, cost, status, latency, request id, IP address and the error message if one occurred). Prompts and responses are not stored.
  • AI Chat: conversations and files you upload are saved to your account so you can come back to them. You can delete them at any time.
  • Requests are processed by the upstream model provider under its own terms. See the Privacy Policy.