AIOS / AI Gateway

Your idea becomes a capability.

Build custom functions, workflows and individual agents. Right where your data lives and your teams work.

AI / FROM CONTEXT TO OUTCOME

Define a task

Start with a clear outcome: which recurring task should your function or agent handle? Describe the required inputs and the objects the work concerns. A concrete example helps you define the scope and gives you a useful basis for assessing the eventual result.

Connect context and tools

Choose the models, skills and tools suited to the task. Connect the context it needs and define which information the agent may use. Inputs, conditions and approvals give the workflow a clear structure before you run it against a concrete example.

Run the function

Run the function where the task arises. AIOS requests the required inputs and carries out the permitted steps. Include the intended approvals in the workflow and assess the result against your original task before using it in the work that follows.

Build on the result

Use the result with its related object or as the starting point for the next task. Once the process proves useful, make it available to relevant objects or include it in a scheduled workflow. Refine its context and inputs using specific experience from real use.

01Define a task

02Connect context and tools

03Run the function

04Build on the result

One module. Many possibilities.

Custom actions for objects and models

Make recurring work available as a function for relevant objects, defining the inputs required for each particular action.

Individual agents with a native runtime

Shape an agent around a specific task and connect its configuration to the capabilities of the native AIOS runtime.

Models, skills and memory

Choose a suitable model, add task-specific skills and decide which memory is useful for the work your agent performs.

Inputs, conditions and loops

Structure the workflow with required inputs, conditions and repeated steps, instead of describing every variation as a separate action.

Scheduled workflows and approvals

Schedule recurring workflows and place approvals where a person should deliberately review the next step before the process continues.

Voice, tools and visible results

Introduce tasks by voice, connect suitable tools and assess the visible results in the context of the work at hand.

Explore the documentation

YOUR PROVIDERS. YOUR CHOICE.

Leading models. Directly connected.

OpenAI, Anthropic, Google Gemini and Kimi come together in one workspace. Use different providers for chat, agents and creative work through the connections authorized for your team.

Your cloud becomes your AI foundation.

AWS Bedrock, Azure OpenAI and Azure Foundry connect AIOS to your cloud infrastructure. Use the models, regions and credentials configured there in your workflows.

More possibilities for every idea.

Hugging Face opens access to the open model ecosystem. fal, Black Forest Labs and Leonardo add specialized image, video and creative models. Choose the connection that fits your outcome.

On your own infrastructure, too.

The AIOS runtime can run suitable models on your own hardware. Local inference keeps model requests within that environment. Models, GPU resources and access permissions are configured for your deployment.

AIOS catalog

OpenAI

Language, image and audio models for agents and generations.

AIOS catalog

Anthropic

Claude for text, analysis and work with tools.

AIOS catalog

Gemini AI

Google models for text and multimodal tasks.

AIOS catalog

Kimi

Moonshot models for language, reasoning and code.

ONE AIOS. MANY MODELS.

A world of models.
Inside your AIOS.

Language, image, video, audio and research. Models in the AIOS catalog are available through your configured providers and runtimes. New models are added continuously.

Explore all connections