Built by Inferable (inferablehq) as an open-source durable execution platform for AI agents and workflows.

Inferable lets developers define long-running AI workflows and agents with structured outputs and human-in-the-loop approval steps that execute inside the user's own infrastructure.

Provided as Node.js and Go SDKs that long-poll to a control plane, deployable behind firewalls or private VPCs without inbound ports.

Introduced its durable workflows feature in 2024 and continues to be maintained on GitHub.

It exists so teams can run reliable, versioned AI agent workflows without exposing internal systems to the public internet.

Requires an Inferable account or control plane, an LLM provider API key, and the Node.js or Go SDK.

Open-source SDK with a hosted control plane, workflow execution happens on your own infrastructure while orchestration metadata is coordinated via Inferable's service.