Platos

Platos is an open-source agent runtime designed for production deployment. It provides everything needed to deploy, monitor, and manage AI agents at scale. Key features include:

• Self-hosted: Deploy in minutes with a single Docker Compose file.
• BYOK (Bring Your Own Key): Use your own provider keys (Anthropic, OpenAI, Google Vertex, etc.) with per-agent routing.
• MCP Gateway: Federate tools from multiple backends through a single endpoint with per-tool ACL.
• Multi-agent clusters: Group agents to share memory and context.
• Memory & Knowledge Graph: Semantic recall, conversation history, and entity extraction.
• Observability: Full tracing, cost attribution, and exportable logs.
• Approval workflows: Gate sensitive tool calls with human-in-the-loop.
• Skills framework: Reusable agent behaviors with version pins.
• Durable background ops: Long-running operations built on trigger.dev.

Platos is ideal for teams building internal agents, customer-facing chatbots, or complex multi-agent systems that require control over infrastructure and data.