NewCost control: a token allowance a month for each model choice — when one runs out, the cheaper one answers.See how it works
memtro
Start for free
Client

Memtro for OpenAI SDKs

Any OpenAI SDK works unchanged: set the base URL and your Memtro key.

Chat completions, the Responses API and embeddings all answer at https://platform.memtro.com/v1, so the Python and JavaScript SDKs need only a base URL and a key. Memtro's options travel in extra_body (or as an extra field in TypeScript), agent progress streams as reasoning_content, and a non-streaming reply carries the routing decision and the tools used in a memtro field next to choices. Tools you declare come back as tool_calls for you to run; Memtro's own run on the server in the same request.

What you get

  • Chat completions with streaming; agent progress as reasoning_content
  • Responses API at /v1/responses for clients that speak only that wire
  • Embeddings on your OpenAI or Google key; model memtro picks the default
  • Client-declared tools, executed by you (not available with claude-code/* models)

How to connect

  1. 01Create an API key on Dashboard → Profile → API keys
  2. 02OpenAI(base_url="https://platform.memtro.com/v1", api_key="sk-memtro-…"), then model="auto" or "auto:agent"

Example queries

  • › Who owns the billing migration?
  • › extra_body={"memtro": {"mode": "agent", "memory": "org"}} for a streamed investigation

Why it is better with memory

Through OpenAI SDKs you get the same memory, connector tools, routing and jobs as every other client, so what one person learns in one tool is available to the whole team in all of them.