Automatic HTTPS provisions TLS certificates for all your sites and keeps them renewed. It also redirects HTTP to HTTPS for you! Caddy uses safe and modern defaults -- no downtime, extra configuration, or separate tooling is required.
<asideclass="tip">Caddy innovated automatic HTTPS technology; we've been doing this since the first day it was feasible in 2015. Caddy's HTTPS automation logic is the most mature and robust in the world.</aside>
- Caddy serves IP addresses and local/internal hostnames over HTTPS using self-signed certificates that are automatically trusted locally (if permitted).
- Examples: `localhost`, `127.0.0.1`
- Caddy serves public DNS names over HTTPS using certificates from a public ACME CA such as [Let's Encrypt](https://letsencrypt.org) or [ZeroSSL](https://zerossl.com).
- Caddy may prompt for a password to install its unique root certificate into your trust store. This happens only once per root; and you can remove it at any time.
- Any client accessing the site without trusting Caddy's root will show security errors.
<asideclass="tip">These are common requirements for any basic production website, not just Caddy. The main difference is to set your DNS records properly <b>before</b> running Caddy so it can provision certificates.</aside>
Because HTTPS utilizes a shared, public infrastructure, you as the server admin should understand the rest of the information on this page so that you can avoid unnecessary problems, troubleshoot them when they occur, and properly configure advanced deployments.
Caddy implicitly activates automatic HTTPS when it knows a domain name (i.e. hostname) or IP address it is serving. There are various ways to tell Caddy your domain/IP, depending on how you run or configure Caddy:
You can [customize or disable automatic HTTPS](/docs/json/apps/http/servers/automatic_https/) if necessary; for example, you can skip certain domain names or disable redirects (for Caddyfile, do this with [global options](/docs/caddyfile/options)).
To serve non-public sites over HTTPS, Caddy generates its own certificate authority (CA) and uses it to sign certificates. The trust chain consists of a root and intermediate certificate. Leaf certificates are signed by the intermediate. They are stored in [Caddy's data directory](/docs/conventions#data-directory) at `pki/authorities/local`.
Caddy's local CA is powered by [Smallstep libraries](https://smallstep.com/certificates/).
Local HTTPS does not use ACME nor does it perform any DNS validation. It works only on the local machine and is trusted only where the CA's root certificate is installed.
### CA Root
The root's private key is uniquely generated using a cryptographically-secure pseudorandom source and persisted to storage with limited permissions. It is loaded into memory only to perform signing tasks, after which it leaves scope to be garbage-collected.
Although Caddy can be configured to sign with the root directly (to support non-compliant clients), this is disabled by default, and the root key is only used to sign intermediates.
The first time a root key is used, Caddy will try to install it into the system's local trust store(s). If it does not have permission to do so, it will prompt for a password. This behavior can be disabled in the configuration if it is not desired.
<asideclass="tip">It is safe to trust Caddy's root certificate on your own machine as long as your computer is not compromised and your unique root key is not leaked.</aside>
After Caddy's root CA is installed, you will see it in your local trust store as "Caddy Local Authority" (unless you've configured a different name). You can uninstall it any time if you wish (the [`caddy untrust`](/docs/command-line#caddy-untrust) command makes this easy).
Note that automatically installing the certificate into the local trust stores is for convenience only and isn't guaranteed to work, especially if containers are being used or if Caddy is being run as an unprivileged system service. Ultimately, if you are relying on internal PKI, it is the system administrator's responsibility to ensure Caddy's root CA is properly added to the necessary trust stores (this is outside the scope of the web server).
To test or experiment with your Caddy configuration, make sure you [change the ACME endpoint](/docs/modules/tls.issuance.acme#ca) to a staging or development URL, otherwise you are likely to hit rate limits which can block your access to HTTPS for up to a week, depending on which rate limit you hit.
One of Caddy's default CAs is [Let's Encrypt](https://letsencrypt.org/), which has a [staging endpoint](https://letsencrypt.org/docs/staging-environment/) that is not subject to the same [rate limits](https://letsencrypt.org/docs/rate-limits/):
Obtaining a publicly-trusted TLS certificate requires validation from a publicly-trusted, third-party authority. These days, this validation process is automated with the [ACME protocol](https://tools.ietf.org/html/rfc8555), and can be performed one of three ways ("challenge types"), described below.
The first two challenge types are enabled by default. If multiple challenges are enabled, Caddy chooses one at random to avoid accidental dependence on a particular challenge. Over time, it learns which challenge type is most successful and will begin to prefer it first, but will fall back to other available challenge types if necessary.
The HTTP challenge performs an authoritative DNS lookup for the candidate hostname's A/AAAA record, then requests a temporary cryptographic resource over port 80 using HTTP. If the CA sees the expected resource, a certificate is issued.
This challenge requires port 80 to be externally accessible. If Caddy cannot listen on port 80, packets from port 80 must be forwarded to Caddy's [HTTP port](/docs/json/apps/http/http_port/).
This challenge is enabled by default and does not require explicit configuration.
### TLS-ALPN challenge
The TLS-ALPN challenge performs an authoritative DNS lookup for the candidate hostname's A/AAAA record, then requests a temporary cryptographic resource over port 443 using a TLS handshake containing special ServerName and ALPN values. If the CA sees the expected resource, a certificate is issued.
This challenge requires port 443 to be externally accessible. If Caddy cannot listen on port 443, packets from port 443 must be forwarded to Caddy's [HTTPS port](/docs/json/apps/http/https_port/).
This challenge is enabled by default and does not require explicit configuration.
### DNS challenge
The DNS challenge performs an authoritative DNS lookup for the candidate hostname's TXT records, and looks for a special TXT record with a certain value. If the CA sees the expected value, a certificate is issued.
This challenge does not require any open ports, and the server requesting a certificate does not need to be externally accessible. However, the DNS challenge requires configuration. Caddy needs to know the credentials to access your domain's DNS provider so it can set (and clear) the special TXT records. If the DNS challenge is enabled, other challenges are disabled by default.
DNS provider support is a community effort. [Learn how to enable the DNS challenge for your provider at our wiki.](https://caddy.community/t/how-to-use-dns-provider-modules-in-caddy-2/8148)
Caddy pioneered a new technology we call **On-Demand TLS**, which dynamically obtains a new certificate during the first TLS handshake that requires it, rather than at config load. Crucially, this does not require specifying the domain names in your configuration ahead of time.
Many businesses rely on this unique feature to scale their TLS deployments at lower cost and without operational headaches when serving tens of thousands of sites.
On-demand TLS is useful if:
- you do not know all the domain names when you start or reload your server,
- domain names might not be properly configured right away (DNS records not yet set),
- you are not in control of the domain names (e.g. they are customer domains).
When on-demand TLS is enabled, you do not need to specify the domain names in your config in order to get certificates for them. Instead, when a TLS handshake is received for a server name (SNI) that Caddy does not yet have a certificate for, the handshake is held while Caddy obtains a certificate to use to complete the handshake. The delay is usually only a few seconds, and only that initial handshake is slow. All future handshakes are fast because certificates are cached and reused, and renewals happen in the background. Future handshakes may trigger maintenance for the certificate to keep it renewed, but this maintenance happens in the background if the certificate hasn't expired yet.
Enabling on-demand TLS happens in [TLS automation policies](/docs/json/apps/tls/automation/policies/) if using the JSON config, or [in site blocks with the `tls` directive](/docs/caddyfile/directives/tls) if using the Caddyfile.
To prevent abuse of this feature, you must configure restrictions. This is done in the [`automation` object of the JSON config](/docs/json/apps/tls/automation/on_demand/), or the [`on_demand_tls` global option](/docs/caddyfile/options#on-demand-tls) of the Caddyfile. Restrictions are "global" and aren't configurable per-site or per-domain. The primary restriction is an "ask" endpoint to which Caddy will send an HTTP request to ask if it has permission to obtain and manage a certificate for the domain in the handshake. This means you will need some internal backend that can, for example, query the accounts table of your database and see if a customer has signed up with that domain name.
You can also configure rate limits as restrictions, though rate limits alone are not a sufficient protection.
Be mindful of how quickly your CA is able to issue certificates. If it takes more than a few seconds, this will negatively impact the user experience (for the first client only).
Due to its deferred nature and potential for abuse (if not mitigated through proper configuration), we recommend enabling on-demand TLS only when your actual use case is described above.
[See our wiki article for more information about using on-demand TLS effectively.](https://caddy.community/t/serving-tens-of-thousands-of-domains-over-https-with-caddy/11179)
Caddy does its best to continue if errors occur with certificate management.
By default, certificate management is performed in the background. This means it will not block startup or slow down your sites. However, it also means that the server will be running even before all certificates are available. Running in the background allows Caddy to retry with exponential backoff over a long period of time.
Here's what happens if there's an error obtaining or renewing a certificate:
1. Caddy retries once after a brief pause just in case it was a fluke
2. Caddy pauses briefly, then switches to the next enabled challenge type
During retries with Let's Encrypt, Caddy switches to their [staging environment](https://letsencrypt.org/docs/staging-environment/) to avoid rate limit concerns. This isn't a perfect strategy, but in general it's helpful.
ACME challenges take at least a few seconds, and internal rate limiting helps mitigate accidental abuse. Caddy uses internal rate limiting in addition to what you or the CA configure so that you can hand Caddy a platter with a million domain names and it will gradually -- but as fast as it can -- obtain certificates for all of them. Caddy's internal rate limit is currently 10 attempts per ACME account per 10 seconds.
To avoid leaking resources, Caddy aborts in-flight tasks (including ACME transactions) when config is changed. While Caddy is capable of handling frequent config reloads, be mindful of operational considerations such as this, and consider batching config changes to reduce reloads and give Caddy a chance to actually finish obtaining certificates in the background.
Caddy is the first (and so far only) server to support fully-redundant, automatic failover to other CAs in the event it cannot successfully get a certificate.
By default, Caddy enables two ACME-compatible CAs: [**Let's Encrypt**](https://letsencrypt.org) and [**ZeroSSL**](https://zerossl.com). If Caddy cannot get a certificate from Let's Encrypt, it will try with ZeroSSL; if both fail, it will backoff and retry again later. In your config, you can customize which issuers Caddy uses to obtain certificates, either universally or for specific names.
Caddy will store public certificates, private keys, and other assets in its [configured storage facility](/docs/json/storage/) (or the default one, if not configured -- see link for details).
**The main thing you need to know using the default config is that the `$HOME` folder must be writeable and persistent.** To help you troubleshoot, Caddy prints its environment variables at startup if the `--environ` flag is specified.
Any Caddy instances that are configured to use the same storage will automatically share those resources and coordinate certificate management as a cluster.
Before attempting any ACME transactions, Caddy will test the configured storage to ensure it is writeable and has sufficient capacity. This helps reduce unnecessary lock contention.
Caddy can obtain and manage wildcard certificates when it is configured to serve a site with a qualifying wildcard name. A site name qualifies for a wildcard if only its left-most domain label is a wildcard. For example, `*.example.com` qualifies, but these do not: `sub.*.example.com`, `foo*.example.com`, `*bar.example.com`, and `*.*.example.com`.
If using the Caddyfile, Caddy takes site names literally with regards to the certificate subject names. In other words, a site defined as `sub.example.com` will cause Caddy to manage a certificate for `sub.example.com`, and a site defined as `*.example.com` will cause Caddy to manage a wildcard certificate for `*.example.com`. You can see this demonstrated on our [Common Caddyfile Patterns](/docs/caddyfile/patterns#wildcard-certificates) page. If you need different behavior, the [JSON config](/docs/json/) gives you more precise control over certificate subjects and site names ("host matchers").
Wildcard certificates represent a wide degree of authority and should only be used when you have so many subdomains that managing individual certificates for them would strain the PKI or cause you to hit CA-enforced rate limits.
**Note:** [Let's Encrypt requires](https://letsencrypt.org/docs/challenge-types/) the [DNS challenge](#dns-challenge) to obtain wildcard certificates.
