Runtimes & models
Choose the right engine for each agent, local or in your cloud.
The engine an agent runs on is a choice, not a lock-in. Padiso separates what an agent is for from what it runs on.
Local and coding agents
Run Claude Code, Codex, Cursor, Gemini, or OpenCode directly on your own machines. Great for code and deep, tool-heavy work.
Hosted models
Use any model through the AI Gateway, including GPT, Claude, Gemini, Llama, Mistral, DeepSeek, Qwen, and Kimi. Pick frontier models where you need them, and fast, cheap models where you do not.
Cloud runtimes
Run agents in your own cloud on Amazon Bedrock AgentCore, Azure AI Foundry, or Google Vertex AI.
Custom adapters
Implement one small interface to run agents on your own infrastructure or a provider Padiso does not ship yet. The runtime owns budgets, retries, and the lifecycle, so adding an engine is a small, isolated change.
Per agent, switchable
Choose a runtime per agent and change it any time without rebuilding anything. A researcher might run on a frontier model while a triage bot runs on something fast and cheap.