Skip to main content
Point your existing OpenAI or Anthropic SDK at SERV and use your SERV_API_KEY — your prompts, tool definitions, and application logic stay the same. Or call the endpoints directly over HTTP. Generate a key at console.openserv.ai.

​
Make your first request

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://inference-api.openserv.ai/v1",
  apiKey: process.env.SERV_API_KEY,
});

const response = await client.chat.completions.create({
  model: "gpt-5.4-mini",
  messages: [
    { role: "system", content: "You are a concise assistant." },
    { role: "user", content: "Hello!" },
  ],
});

console.log(response.choices[0].message.content);

import Anthropic from "@anthropic-ai/sdk";

const client = new Anthropic({
  baseURL: "https://inference-api.openserv.ai",
  authToken: process.env.SERV_API_KEY,
});

const message = await client.messages.create({
  model: "claude-haiku-4.5",
  max_tokens: 1024,
  system: "You answer in one sentence.",
  messages: [{ role: "user", content: "Hello!" }],
});

console.log(message.content[0].text);
That completes the integration. For Python, the Vercel AI SDK, LangChain, and other clients, see SDK Migration.

​
Call the endpoints directly

For raw HTTP, SERV exposes three endpoints under https://inference-api.openserv.ai. 
curl https://inference-api.openserv.ai/v1/chat/completions \
  -H "Authorization: Bearer $SERV_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.4-mini",
    "messages": [
      {"role": "system", "content": "You are a concise assistant."},
      {"role": "user", "content": "Hello!"}
    ]
  }'

curl https://inference-api.openserv.ai/v1/responses \
  -H "Authorization: Bearer $SERV_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.4",
    "instructions": "You are a careful reasoner.",
    "input": "Hello!"
  }'

curl https://inference-api.openserv.ai/v1/messages \
  -H "Authorization: Bearer $SERV_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-haiku-4.5",
    "max_tokens": 1024,
    "system": "You answer in one sentence.",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

​
Things to know

A few SERV-specific behaviors are worth knowing before your first integration. Full detail is in SDK Integration.
  • A system prompt is required. Every request needs a system message, instructions field, or developer message. Requests without one are rejected.
  • The base URL differs by SDK. The OpenAI SDK takes /v1 in the base URL; the Anthropic SDK does not — it appends /v1/messages itself.
  • The auth field differs by SDK. OpenAI uses apiKey; the Anthropic SDK uses authToken.

​
Try it in the Playground

Compare any model with and without SERV Reasoning side by side on the same prompt before you commit to an integration. → console.openserv.ai/playground

​
Next steps

I'm migrating my project from [OpenAI / Anthropic / other provider] to SERV Reasoning.
SERV is wire-compatible with the OpenAI Chat Completions API and the Anthropic
Messages API. Please refactor as follows:

1. Set the client baseURL:
   • OpenAI-shape clients   -> "https://inference-api.openserv.ai/v1"  (WITH /v1)
   • Anthropic-shape client -> "https://inference-api.openserv.ai"    (WITHOUT /v1)

2. Read the API key from SERV_API_KEY.
   • OpenAI SDK: use the `apiKey` constructor field.
   • Anthropic SDK: use the `authToken` constructor field, not `apiKey`.

3. Leave model IDs as-is if they are already SERV-supported. The current public
   catalog is:
     OpenAI-shape: gpt-5.5, gpt-5.4, gpt-5.4-mini, gpt-5.4-nano,
                   o3, o3-mini, o3-pro, o4-mini,
                   gemini-flash-latest, gemini-pro-latest,
                   gemma-4-26b-a4b, gemma-4-31b,
                   grok-4.3, grok-4.20,
                   qwen3.6-flash, qwen3.6-max-preview,
                   deepseek-v4-pro, deepseek-v4-flash
     Anthropic-shape: claude-opus-4.6, claude-sonnet-4.6, claude-haiku-4.5
   If you encounter a non-supported ID, flag it for me to pick a replacement.

4. SERV requires a system prompt on every request. Audit every call site. If any
   request lacks a system message / instructions / developer message, add a
   default ("You are a helpful assistant." is fine) and flag it for me to refine.

5. If any code passes `reasoning_effort: "minimal"`, change to `"low"`. SERV's
   allowed values are: none | low | medium | high.

6. Leave all prompts, tool/function definitions, message structures, streaming
   logic, and business logic UNCHANGED.

Produce a diff for every call site, and flag any provider-specific features
(prompt caching, assistants API, etc.) for manual review.

Models

The full model catalog with pricing and context windows.

SDK Integration

Endpoint details and the parameter map.

SDK Migration

Migrate Python, Vercel AI SDK, LangChain, and raw fetch clients.

Why SERV Reasoning

What you get over calling model APIs directly.