Use OpenAI SDK with Manifest

Home · Agents · OpenAI SDK

— / 30 days

App AI SDK

The OpenAI Python and TypeScript SDKs talk to any OpenAI-compatible endpoint. Point them at Manifest by setting OPENAI_BASE_URL and every chat completion, response, tool call, or Agents SDK loop routes through the cheapest capable model.

— models routed through Manifest over the last 30 days.

Why use the OpenAI SDK with Manifest

Chatty workloads (Agents SDK loops, tool retries, chain-of-thought reasoning) burn tokens fast. Manifest intercepts each call, picks the right model, and keeps your code identical. Same SDK, same shapes, lower bill.

One env var change: OPENAI_BASE_URL
Slash token spend by 50–70% on chatty agent loops
Works for the OpenAI SDK and the OpenAI Agents SDK
Tool calling, structured outputs, and streaming all pass through
Per-app spend caps and audit logs in the dashboard
Open source: inspect and self-host the router

Setup in 60 seconds

Set two environment variables and the OpenAI SDK picks them up on the next run, no code changes required.

Point the SDK at Manifest by setting these env vars in your shell or .env file:

export OPENAI_BASE_URL="https://api.manifest.build/v1"
export OPENAI_API_KEY="mnfst_YOUR_KEY"

Then use the SDK exactly as you would normally. Just point at "auto" to let Manifest pick the model:

from openai import OpenAI

client = OpenAI()

response = client.chat.completions.create(
    model="auto",
    messages=[{"role": "user", "content": "Hello"}],
)

import OpenAI from "openai";

const client = new OpenAI();

const response = await client.chat.completions.create({
  model: "auto",
  messages: [{ role: "user", content: "Hello" }],
});

OPENAI_BASE_URL https://api.manifest.build/v1

OPENAI_API_KEY mnfst_YOUR_KEY

model auto

OpenAI SDK tokens processed on Manifest

Models in use

# Model	Auth type	Tokens / last 30 days