Inception

Inception builds Mercury, a family of diffusion large language models (dLLMs) that refine all tokens in parallel instead of generating them left-to-right, making them very fast. Inception serves the models through an OpenAI-compatible API, so you drive them through Agno like any other OpenAI-compatible provider. The Inception class defaults to mercury-2 and points at https://api.inceptionlabs.ai/v1.

Installation

uv pip install -U openai agno

Authentication

Create an account at the Inception Platform.
Open the dashboard and go to API Keys.
Create a key and set your INCEPTION_API_KEY environment variable.

export INCEPTION_API_KEY=***

setx INCEPTION_API_KEY ***

Example

Use Inception with your Agent:

from agno.agent import Agent
from agno.models.inception import Inception

agent = Agent(model=Inception(id="mercury-2"), markdown=True)

# Print the response in the terminal
agent.print_response("Share a 2 sentence horror story.")

View more examples here.

Parameters

Parameter	Type	Default	Description
`id`	`str`	`"mercury-2"`	The id of the Mercury model to use
`name`	`str`	`"Inception"`	The name of the model
`provider`	`str`	`"InceptionLabs"`	The provider of the model
`api_key`	`Optional[str]`	`None`	The API key for Inception (defaults to `INCEPTION_API_KEY` env var)
`base_url`	`str`	`"https://api.inceptionlabs.ai/v1"`	The base URL for the Inception API

Inception extends the OpenAI-compatible interface and supports most parameters from the OpenAI model. Note: Inception’s OpenAI-compatible endpoint does not advertise native json_schema structured outputs, so supports_native_structured_outputs is set to False. Use use_json_mode=True for structured output. To hit a different host (private deployment, regional endpoint), pass base_url:

Inception(id="mercury-2", base_url="https://your-host.example.com/v1")

Available Models

Model id	Notes
`mercury-2`	Flagship reasoning dLLM. Tunable reasoning depth, 128K context, native tool use, JSON output.
`mercury-coder-small`	Coding-focused variant for latency-sensitive code workflows.

The original mercury model is only available to accounts created before February 24, 2026. New accounts should use mercury-2 (or the coder variants) instead.

Get Started

Basics

Advanced

Production

Providers

Other

Additional Resources

Installation

Authentication

Example

Parameters

Available Models

​Installation

​Authentication

​Example

​Parameters

​Available Models

Installation

Authentication

Example

Parameters

Available Models