AI agents for business

What an agent needs from its infrastructure

An agent is a loop of model → tool → model, and most of what breaks in production breaks at the model-access layer: a quota runs out, a provider errors, a step turns out to need a cheaper or a stronger model. Here is the part of that our API covers.

The models agents are built on

Price per million input tokens, cheapest first. The table is read from the live catalog, so it cannot drift from what your key can actually call.

What gets built on this

Four shapes companies usually start with — a sense of the scope of the work rather than an exhaustive list.

How to connect an agent

Integration means changing a base URL and a key in code you have already written. No SDK of ours is required: we speak the OpenAI and Anthropic formats, so the agent framework, the tools and the prompts stay as they are.

1. 1

   Get a key

   Sign up, top up the balance and create an API key in the dashboard. One key covers every model — chat and media alike.

2. 2

   Point the base URL at us

   Give the OpenAI or Anthropic client our address and your key. The rest of the agent — tools, memory, call chains — is untouched.

3. 3

   Route models per step

   Name the model each step needs in the `model` field. Spend per model and per key shows up in the dashboard statistics.

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["API_KEY"],
    base_url="https://api.apihubs.ru/api/v1",
)

response = client.chat.completions.create(
    model="deepseek-v4-flash",
    messages=[{"role": "user", "content": "..."}],
    tools=tools,
)

curl https://api.apihubs.ru/api/v1/chat/completions \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4-flash",
    "messages": [{"role": "user", "content": "..."}]
  }'

Frequently asked questions