NyxId

Open-source connectivity gateway for AI agents. Point Claude Code, Cursor, or n8n at NyxID and they can reach any API — public, internal, or on localhost — without ever seeing a raw key. Credentials are injected at proxy time. NAT traversal reaches private services with no VPN. Any OpenAPI spec auto-wraps into MCP tools. Per-agent scoped keys, phone approvals, instant revoke. Apache-2.0. TL;DR

1. Give every AI its own key Claude, Cursor, Codex, your weekend side-project bot — each gets a scoped NyxID key instead of your real OpenAI/GitHub/Anthropic keys. If one leaks or goes rogue, revoke that one key. See exactly which agent spent what.
2. One place to plug in every API your AI touches Stop pasting OpenAI, GitHub, Slack, and Notion keys into every agent you run. Point them at NyxID and it injects the right credentials per call. Swap providers without touching agent config.
3. Or keep credentials fully on-prem Can't put secrets in someone else's cloud? Install the optional node agent inside your VPC, server, or laptop and your real credentials never leave it. The control plane still brokers the request, but injection happens locally, on your side of the wire. Opt in for data residency, tighter blast radius, or compliance sign-off — skip it when you don't need them.
4. Approve big AI actions from your phone Set rules so the sensitive stuff — prod deploys, large charges, data exports — pings your phone first. Tap approve or deny before the AI runs it. A bank confirmation, for your agents.
5. One identity for you and your AI Sign in with NyxID (email, Google, GitHub, MFA) and the token you get is the identity your agents use to reach every API. No separate secrets vault, no separate login system. Connect OpenAI, GitHub, or your internal services once, and every AI tool you run inherits your permissions, budgets, and audit trail through that single Nyx token.