Networking

A Sprite runs your code in the cloud, but it’s meant to feel close. Traffic gets in two ways, an always-on HTTPS URL and a proxy that maps a remote port onto your laptop. A network policy governs what the Sprite can reach on its way out. This page is the model for both directions: how you reach in, and what a Sprite can reach going out. The exact commands live on the pages linked throughout.

Reaching a Sprite

Inbound traffic arrives one of two ways, and which you pick depends on whether you want a public endpoint or a local one.

• The Sprite URL is how you reach an app over the Internet: a webhook target, a shared demo, an API. Every Sprite has one at https://<sprite-name>-<org-id>.sprites.app/ (the org ID is a short generated identifier; sprite info prints your exact URL), and it routes HTTPS to the Sprite’s HTTP service. It’s always on: no CLI required.
• sprite proxy is how you make a Sprite feel local: it maps a remote port onto your machine, so you can point a database client or a browser at a service running in the Sprite.

# Show the Sprite's URL and its auth setting
sprite info

# Map the Sprite's port 5432 onto localhost:5432 for a psql client
sprite proxy 5432

	Sprite URL	sprite proxy
Protocol	HTTP(S) only	Any TCP
Reach	Public Internet	Your machine only
Needs the CLI running	No	Yes
Ports	One (the HTTP service)	As many as you forward

sprite proxy also remaps ports (sprite proxy 3001:3000) and tunnels stdin and stdout for things like SSH (sprite proxy -W :22). And you often don’t reach for it at all: when you run sprite exec and your command opens a listening port, the CLI forwards that port to your laptop automatically. For the full command surface and the SSH-over-proxy setup, see Working with Sprites and the CLI Commands reference.

The HTTP service the URL routes to, including how it wakes on an incoming request and how to move it off the default port, is covered in Services.

URL authentication

A Sprite URL is private by default. It’s reachable only by members of your org, through the browser or with an org token, so standing up a service doesn’t put it on the open Internet by accident.

Make it public when you actually want that behavior, for example, a webhook that needs to be hit without a token, a demo you’re sharing, or putting something quick on the Internet:

# Anyone with the URL can reach it
sprite config update --url-auth public

# Back to org-only (the default)
sprite config update --url-auth sprite

A public URL is on the open Internet

Public means no authentication: anyone with the URL reaches whatever the HTTP service serves. Don’t make a URL public if it exposes secrets, tokens, or internal endpoints, and put your own auth in front of anything real. Flip it back to sprite when you’re done.

Reaching out

By default, a Sprite’s outbound is unrestricted: it can resolve and reach any domain. Egress can be tightened with a network policy, a DNS-based allowlist that decides which domains a Sprite is allowed to reach. Applying one is opt-in and done from outside the Sprite. Once a policy is in force, a request to an allowed domain works normally, while a request to one that isn’t gets a DNS REFUSED and fails fast rather than hanging.

The policy is a set of rules, read-only inside the Sprite at /.sprite/policy/network.json:

{
  "rules": [
    { "include": "defaults" },
    { "domain": "example.com", "action": "allow" },
    { "domain": "*.example.com", "action": "allow" },
    { "domain": "blocked.com", "action": "deny" }
  ]
}

• { "include": "defaults" } pulls in the common development domains, GitHub, npm, PyPI, Docker Hub, and the major AI APIs among them, so package installs and model calls work without listing every host yourself.
• Domain rules match an exact host (example.com), a subdomain wildcard (*.example.com), or everything (*). More specific rules win: an exact match beats a subdomain wildcard, which beats the global wildcard.
• { "rules": [] } means no enforcement. The Sprite runs unrestricted.

When a policy is enforced, a few behaviors follow from its DNS-based design:

• Raw IP connections are blocked unless the IP was resolved from an allowed domain. You can’t route around the allowlist by dialing an address directly.
• Private IPs are always blocked, so a Sprite can’t reach into private network ranges.
• Changes reload live. When a policy tightens, existing connections to newly-blocked domains are dropped rather than left open.

The policy is read-only from inside: a Sprite can’t rewrite its own egress rules. Changes are made from outside through the Sprites API. To test what the current policy allows, resolve a domain and watch for REFUSED:

dig github.com        # an allowed domain resolves
dig blocked.com       # a denied domain returns REFUSED

For calling external APIs without handing the Sprite a long-lived credential, Connectors route the request through a gateway that holds the token for you. That’s a separate mechanism from the egress policy: the policy decides what a Sprite may reach, Connectors decide what it may reach as.

That’s the shape of networking on Sprites: the URL and proxy control how you reach a Sprite, the egress policy controls how far it reaches back. You’re in control of both, so it’s worth opening each only as far as your work actually needs.

Related documentation

Working with Sprites

Port forwarding, URL auth, and SSH over the proxy in practice

Services

The HTTP service the URL routes to, and waking on a request

Connectors

Reach external APIs with credentials brokered for you

CLI Commands

Full reference for the URL and proxy commands