Questions developers ask before they trust infrastructure

Persistent infrastructure for AI coding agents. It wraps Claude Code, Codex, Antigravity, Grok Build, OpenCode, and shell sessions behind a self-hosted gateway.

Yes. OpenDray runs entirely on your own machine or server as a single Go binary. Sessions, credentials, memory vectors, project files, and the audit trail stay on infrastructure you control, with no telemetry.

The sessions, credentials, memory vectors, project files, and audit trail stay on infrastructure you control.

OpenDray spawns real PTY subprocesses and shares host state such as project files, SSH agent, and CLI credentials. Production Docker isolation fights that model.

Claude Code, Codex, Antigravity, Grok Build, OpenCode, and plain shell are first-class targets, with provider updates managed by the binary.

When a session needs attention, OpenDray can notify a channel. Your reply is routed back into the right session.

Yes. OpenDray exposes REST and WebSocket APIs with scoped keys and audit logging.

The gateway is self-hosted and local-first. Memory embeddings can run with ONNX, Ollama, or LM Studio.

A normal CLI dies with your local context. OpenDray keeps the agent session running on server infrastructure and lets you reconnect from anywhere.

Jev (TypeSafe) is a decision CLI that returns calibrated JSON probabilities and a verdict instead of prose. You set one API key in a [jev] config section and OpenDray injects it into every session, so any provider's agent can call the jev CLI to classify, verify, route, screen, match, or rank with no per-provider setup.

Yes. OpenDray is open source under the Apache 2.0 license, free to self-host, with no telemetry and no required cloud account.