Problem solved

Developers and AI engineers often struggle to build autonomous agents that maintain state, safely execute code, and orchestrate multi-step workflows without heavy custom infrastructure. aios solves this by providing a ready-to-use, open-source runtime that handles event-driven orchestration, persistent sessions via PostgreSQL, and secure Docker sandboxes, all while offering the flexibility to choose any LLM model via LiteLLM. This reduces boilerplate and accelerates the development of reliable, production-ready AI agents.

Use cases

• Building a customer support agent that retains conversation history and executes safe tool calls within a sandbox
• Developing an automated code reviewer that analyzes pull requests using an LLM and runs tests in isolated Docker containers
• Creating a research assistant that chains multiple LLM queries to gather, summarize, and store information across sessions
• Implementing a data pipeline agent that schedules and monitors ETL jobs with state tracked in PostgreSQL
• Deploying a multi-agent system where agents collaborate through events and share session data

Audience

• Developers
• ML Engineers
• Researchers

Pros

• Open-source and free to use
• Supports any LiteLLM model for broad LLM compatibility
• Docker sandboxing enhances security for code execution
• Postgres-backed sessions provide robust state management
• Event-driven architecture enables complex agent workflows

Cons

• Requires PostgreSQL and Docker setup, adding operational overhead
• Documentation may be sparse or evolving as a newer project
• Dependency on external services like Postgres could complicate simple use cases