Remote Executors overview

Some of the things you want an agent to do can only happen inside your own network. A database that isn’t on the internet. An internal API behind a VPN. A GPU box in your own rack. The hosted qRaptor runtime can’t reach any of that, and opening it up so it could is usually not an option.

A remote executor solves this by turning the problem around. You run a small agent — qRemoteX — on your own infrastructure. It dials out to qRaptor and holds the connection open. When one of your agents needs to run code or call a tool that belongs inside your network, qRaptor sends the work down that existing connection, your executor runs it, and sends the result back.

What you can run remotely

Three things in qRaptor can be pointed at an executor:

Everything else — the LLM calls, the agent’s reasoning, memory, deployments — keeps running on the platform as usual. An executor handles the specific steps you point at it, not the whole agent.

What runs where

Your infrastructure                        qRaptor
┌──────────────────────┐                  ┌──────────────────────┐
│  qRemoteX            │ ──── outbound ──▶│  Executor gateway    │
│                      │      WSS         │                      │
│  runs your code      │◀──── work ───────│  decides what runs   │
│  reads your database │───── results ───▶│  remotely            │
└──────────┬───────────┘                  └──────────────────────┘
           │
           ▼
  Private DBs · internal APIs · files · GPUs

The executor is deliberately simple. It doesn’t hold credentials for your qRaptor project beyond its own API key, and it never reads your secret store — any secrets a tool needs are resolved on the platform before the work is sent. See Security & networking.

What you need

• Pro Pro plan or above. Remote executors aren’t available on Free or Starter.
• A machine or container host that can make outbound HTTPS connections to qraptor.app. It does not need a public IP or an inbound rule.
• Docker is the simplest way to run it. There’s also a pip package, a Helm chart, and a custom-image path if you need your own dependencies.

How this section is organised

Setting one up — follow these three in order if you’re the person installing the executor.

• Turn on remote execution — activate the gateway for your project and get your connection URL.
• Add an executor — name it, decide how it should behave, generate its API key.
• Install & connect — run the agent and watch it come online.

Sending work to it — start here if an executor is already running and you’re building the agent.

• Run work on an executor — how qRaptor decides where a step runs, and how to point code nodes, tools and MCP servers at your infrastructure.

Running it day to day

• Manage executors · API keys · Monitor executions · Troubleshooting · Security & networking

Next

• Turn on remote execution — the first step.