Works with any AI agent that calls an LLM.

InfraNistic exposes a simple HTTP POST endpoint. Any agent or framework that can make an HTTP call can use it:

import requests

response = requests.post(
    "https://your-infranistic-endpoint/",
    json={"query": your_prompt},
    timeout=600
)
answer = response.json()["response"]

For agents with configurable API endpoints, point the base URL at your InfraNistic endpoint and use query or prompt as the input field.

Works with any model pair available on Amazon Bedrock. Default configuration uses Claude Haiku and Sonnet. Custom model pairs configurable via environment variables — point InfraNistic at any two Bedrock models and it routes between them automatically.

For agentic loops with dynamic or stateful queries — tool calls, status checks, live data — pass "no_cache": true to ensure every call hits the model fresh:

response = requests.post(
    "https://your-infranistic-endpoint/",
    json={"query": your_prompt, "no_cache": True},
    timeout=600
)
answer = response.json()["response"]

InfraNistic acts as the economic brain for any AI agent — ensuring every call is cost-effective and accurate. No changes to the agents themselves.