--- title: "Deploy orca" description: "Remote Orca server: an agent development environment you pair devices to" category: "AI/ML" url: https://railway.com/deploy/orca --- # Deploy orca Remote Orca server: an agent development environment you pair devices to **[Deploy orca on Railway](https://railway.com/template/orca)** Machine-readable deploy manifest (JSON, validated by TemplateCI): https://railway.com/deploy/orca/manifest.json - **Creator:** Paul Zarudnev's Projects - **Category:** AI/ML ## Template content ### orca - **Source:** https://github.com/RockinPaul/orca_railway_template - **Health check:** / - **Public domain:** Yes ## Documentation # Deploy and Host Orca on Railway [Orca](https://github.com/stablyai/orca) is an agent development environment: worktrees, terminals and long-running coding-agent sessions in one workspace. This template deploys a **Remote Orca Server** — Orca 1.4.203 running headless on Railway — so that workspace lives on a machine that stays awake, and you pair your desktop, browser or phone to it. ## About Hosting Orca Orca is an Electron desktop application, and upstream ships desktop packages rather than a container image. This template installs the published Debian package, verified against a pinned digest, and runs its documented `orca serve` mode. Running Electron headless on a platform container takes four things that are easy to get wrong: a virtual display, one shared library the package does not declare, a non-root user, and a disabled Chromium sandbox — all of them are handled in the image. There is nothing to fill in before deploying. Orca mints a pairing identity on first boot and prints the pairing URL to the deploy log, and a volume keeps that identity — along with your repositories, agent accounts and configuration — across redeploys. ## Common Use Cases - Keep a coding agent working on a long task in a workspace you can check on from anywhere. - Drive worktrees, terminals and agent sessions from a tablet or phone, with the work running on a machine that does not sleep or close its lid. - Give a project a persistent development environment separate from your laptop. ## Dependencies for Orca Hosting - A volume for the workspace, which the template creates. - Nothing else. No database, no external service, no account to register. ### Deployment Dependencies - [Orca](https://github.com/stablyai/orca) — the upstream project (MIT). - [Orca documentation](https://www.onorca.dev/docs/ways-to-run) — including Remote Orca Servers. - [Template repository](https://github.com/RockinPaul/orca_railway_template) (MIT). ### Implementation Details One service builds from this template's repository: - **orca** — `debian:13-slim` with `orca-ide` 1.4.203 installed from the upstream `.deb`, checked against a pinned SHA-256 for each architecture. It serves HTTP and WebSocket on port 8080, answers the platform health check on `/`, and mounts a volume at `/data`. The service's home directory is on that volume, set in `/etc/passwd` rather than only in the environment, because the entrypoint drops privileges and that resets `HOME` from the passwd entry. This is what makes the pairing identity stable: without it every redeploy mints a new one and the URL you saved stops working. The address clients dial is derived at boot as `wss://$RAILWAY_PUBLIC_DOMAIN`. Orca advertises whatever address it is given verbatim, and a bare hostname produces a `ws://host:8080` URL that Railway's edge — TLS on 443 only — cannot serve. **First run:** open the service's deploy logs and find the `Web client URL` and `Pairing URL` lines. Open the web client in a browser, or paste the pairing URL into Orca desktop under *Settings → Remote Orca Servers → Add Server*. Install the agents you want from a shell on the service (`orca account add --agent claude`), because the remote client disables *Add account*. **The pairing URL is the credential for this deployment.** Anyone who has it controls the workspace, and agents in it run shell commands, so treat it like a password. The transport is fail-closed: the WebSocket upgrade is anonymous by necessity, and the device token is then presented inside an encrypted channel, with a bad token closing the connection. Note that upstream marks Remote Orca Servers **beta** and scopes them to a private network path such as a tailnet or LAN; a Railway domain is public, so this runs the feature further out than upstream intends. Two limits are worth knowing. Running sessions end when the service redeploys, because a redeploy replaces the container — files on the volume survive, terminals and agent sessions do not. And upstream releases every day or two, so the pinned version moves quickly; the template repository documents how to bump it. ## Why Deploy Orca on Railway? Railway is a singular platform to deploy your infrastructure stack. Railway will host your infrastructure so you don't have to deal with configuration, while allowing you to vertically and horizontally scale it. By deploying Orca on Railway, you are one step closer to supporting a complete full-stack application with minimal burden. Host your servers, databases, AI agents, and more on Railway. ## Similar templates - [Chat Chat](https://railway.com/deploy/-WWW5r) — Chat Chat, your own unified chat and search to AI platform. - [stella](https://railway.com/deploy/stella) — Self-host stella with web, API, Postgres, Redis, and object storage. - [Hermes Agent | OpenClaw Alternative with Dashboard](https://railway.com/deploy/hermes-agent-or-openclaw-alternative-wit) — Self-Hosted Hermes AI Agent for Telegram, Discord & Slack Open this page in a browser: https://railway.com/deploy/orca