For coding agents

Software written by agents
is run by agents.

nyxory is the agent your coding agent calls in when it needs to ship. Hundreds of deploys, every error, every fix — already in its head. Your agent doesn’t guess.

Connect your agent Book a demo
Works with Claude Code, Cursor, Codex, and Lovable

What connecting does

Connect once. It runs the rest.

Your coding agent connects to nyxory over MCP. From your chat it ships the project live, keeps it running, and hands control straight back to you.

See how to connect
the connection you prompt · it runs · you stay in control
STAYS IN YOUR CHAT YOUR CODING AGENT Claude Code · Cursor · Codex · Lovable CONNECT OVER MCP NYXORY RUNS IT YOUR PROJECT, KEPT RUNNING SHIP WATCH FIX ON THE RECORD LIVE URL, TLS yourproject.nyxory.app live, and staying live

Schematic. Your agent drives the tools; nyxory runs the project.

Connect in one command

Paste it. Sign in. Ship.

One command installs the nyx CLI, signs you in through the browser, and wires every coding agent on your machine. No token to paste.

brew install nyxory/tap/nyx && nyx setup

Detected macOS. nyx handles the rest — sign-in, setup, every client on your machine.

  1. Step 1 Paste & run The command above, in any terminal. Your browser opens to sign you in.
  2. Step 2 Restart your agent Close and reopen Claude Code, Cursor, or your editor so the connector loads.
  3. Step 3 Tell it to ship “Deploy this to nyxory.” That’s the whole command.
What your agent gets

A tool surface built for agents.

Not a dashboard for humans. The full catalog is machine-readable; here is the shape of it.

Deploy from your repo

Clone, build, and ship from a GitHub or GitLab URL. Zero config — your agent picks the right build for the project.

Projects, secrets, custom domains

The resources your apps need: env vars, secret values, hostnames with automatic TLS. One resource pool, split freely across projects.

Live status and logs

Stream logs, watch your app come up, read progress as it happens. Your agent debugs like an SRE, not a guesser.

Built to diagnose

When your app misbehaves, your agent pulls its real conditions, history, and logs in one call, finds the cause, and ships the fix. The truth, not a guess.

Coding agents

Pick your agent. Connect once.

The one command above connects all of them at once. Only need one? Every agent below has its own one-click or one-command path — all OAuth, no tokens.

Claude Code

CLI

Run in your terminal, restart the session, approve OAuth on first use.

claude mcp add --transport http --scope user nyxory https://api.nyxory.com/mcp
Full Claude Code guide

Cursor

IDE + CLI agent

One click adds the nyxory connector; sign-in happens in your browser.

Add to Cursor

Codex

OpenAI CLI + app

The one command above wires Codex — config and browser sign-in included.

Manual setup

Add this to ~/.codex/config.toml (app + CLI share it):

[mcp_servers.nyxory] url = "https://api.nyxory.com/mcp" enabled = true

Then sign in explicitly — Codex never prompts on its own:

codex mcp login nyxory

Claude Desktop

macOS / Windows app

Download the extension, double-click to install, sign in on first use.

Download nyxory.mcpb

VS Code

GitHub Copilot Chat

One click registers the connector with Copilot agent mode.

Add to VS Code

Any MCP-capable agent

Streamable HTTP

Point your MCP config at this URL as nyxory — OAuth on first connect.

https://api.nyxory.com/mcp
Why nyxory

Every coding agent needs nyxory.

Devs have DevOps. Agents have nyxory.

The nyxory sub-agent keeps prod in flow.

Think of nyxory as the seasoned teammate your coding agent calls in. Hundreds of deploys, hundreds of bugs, hundreds of “wait, why isn’t this working” — already solved, packed into our sub-agent.

When something doesn’t go to plan, your agent doesn’t guess. nyxory hands it the next move, because we have already been in that exact spot a hundred times. Other platforms hand your agent an API. We hand it a sub-agent that runs production.

European hosting OAuth, no tokens Isolated per project Your code stays in your Git
Troubleshooting & support

Stuck? Three quick fixes.

Most connect snags are one of these. Still stuck, a human is one message away.

Your agent doesn't see nyxory

The connector loads when your agent starts. Close and reopen Claude Code, Cursor, or your editor after connecting, then ask it to list your projects.

Sign-in didn't go through

Connecting is browser OAuth — no token to paste. Run the connect command again and approve the tab your browser opens. One approval covers every tool.

It can't reach a private repo

Public repos clone right away. For a private GitHub repo, add the nyxory GitHub App when your agent shares the link — it clones over a short-lived token, no keys to hand over.

Your first project goes live today.

Connect your agent, or start in the console — free to begin, EU hosted.

Ask any AI about nyxory