OpenSink

The product is a memory and control layer for AI agents that records activity, manages human approvals, and provides a real-time timeline of execution, addressing lack of visibility, scattered outputs, and inability to pause and resume agents in production. It sells to developers and operations teams in B2B companies running AI agents in production where outputs must persist beyond a single run. It is positioned as a framework-agnostic control plane that works with all major AI frameworks and providers and stores structured data as a system of record rather than logs or traces.

Key features

• Durable resumable agent sessions
• Human-in-the-loop approvals
• Real-time execution timeline
• Durable state storage
• Sinks for durable outputs
• Structured items in sinks
• Scheduled jobs via HTTP triggers
• Agent configurations management
• Flat explicit schema-driven APIs
• MCP-compatible API access