Tuning Engines Cookbook

Tuning Engines exposes an OpenAI-compatible endpoint for teams that want Agno agents to run through a governed AI control plane. Agno owns the agent behavior, tools, memory, and orchestration. Tuning Engines centralizes model access, policy checks, audit logs, traces, and usage/cost reporting.

1. Create an inference key

Create a Tuning Engines inference key and enable the model alias you want the agent to use.

2. Export environment variables

export TUNING_ENGINES_API_KEY=sk-te-your-inference-key
export TUNING_ENGINES_MODEL=gpt-4o

If you run Tuning Engines behind a custom host, also set:

export TUNING_ENGINES_BASE_URL=https://your-host.example.com/v1

3. Install libraries

uv pip install -U agno openai

4. Run the example

python cookbook/90_models/tuning_engines/basic.py

The example uses the dedicated TuningEngines model provider:

from agno.agent import Agent
from agno.models.tuning_engines import TuningEngines

agent = Agent(model=TuningEngines(id="gpt-4o"))