# Address IPFS on the web

This page describes how to address a node in the IPFS network. Clients that support the IPFS protocol can ignore HTTP details and retrieve data natively, while those that don't can fetch the resource from HTTP server at ipfs.io gateway, as long as they have the content identifier (CID). When ipfs.io or any other public gateway (opens new window) goes down, IPFS aware clients will still be able to fetch the content from the IPFS network as long as at least one node still provides the data behind the CID to the network:

Addresses using a gateway use the following form, where <gateway> is the gateway address, and <CID> is the content identifier


For example:


A self-hosted local gateway (opens new window) can also be used, instead of ipfs.io.

# IPFS addressing in brief

In IPFS, content addresses are path-like; that is, the addresses are components separated by slashes. The first component is the protocol, which tells you how to interpret everything after it.

Content referenced by a hash may have named links. For example, a Git commit has a link named parent, which is really just a pointer to the hash of another Git commit. Components in an IPFS address after the CID are the named links.

Since content addresses aren’t URLs, using them in a web browser requires reformatting. The options for this are:

  1. A path gateway

  2. A subdomain gateway, for hosting websites with origin isolation. This is more secure, but harder to set up.

  3. Native protocol handlers, when you don't want to hard-code a specific HTTP gateway in the URI:


# HTTP gateways

HTTP gateways allow tools that "speak" HTTP but do not speak "IPFS" to communicate. They are the first stage of the upgrade path for the web. More information about IPFS Gateways.

One downside of HTTP gateways is centralization. Location-based addressing of a gateway depends on both DNS and HTTPS/TLS, which relies on trust in certificate authorities (opens new window) (CAs) and public key infrastructure (opens new window) (PKI). In the long term, these issues should be mitigated by the use of opportunistic protocol upgrade schemes.

# Protocol upgrade

Long term, deserialized responses returned by a public HTTP gateway are used only as a fallback when no native implementation of IPFS is available.

IPFS clients, user agents, tools and extensions should detect CIDs in URLs, DNSLinks or IPFS content paths and resolve them directly over the IPFS protocol. This ensures that the retrieved data matches the expected hash.

Examples of user agents that support IPFS natively are:

# Path gateway

A path gateway is the most basic scheme. In this scheme, a URL path used for content addressing is effectively a resource name without a canonical location. The HTTP server provides the location part, which makes it possible for browsers to interpret an IPFS content path relative to the current server and work without need for any conversion. Given a gateway host address (i.e. ipfs.io), and a path to the resource, (i.e /path/to/resource), a CID (<cid>), IPNS ID (<ipnsid>) or DNSLink (<dnslink>) can all be used.

Using a CID Using IPNS Using DNSLink


Given a CID <cid>, a URL path can be constructed as follows:





Given an IPNS name (opens new window) <ipns-name>, a URL path can be constructed as follows:




Given a DNS name with a DNSLink (opens new window) text record <dnslink>, a URL path can be constructed as follows:





In this scheme, all pages share a single origin (opens new window). As such, this type of gateway should only be used when site isolation does not matter. Examples include static content without cookies, local storage, or APIs that require user permission.

When in doubt, use a subdomain gateway.

# Subdomain gateway

When origin-based security (opens new window) is needed, a CIDv1 in a case-insensitive encoding such as Base32 or Base36 should be used in the subdomain:




# Native support in Kubo

Kubo (opens new window) provides native support for subdomain gateways on hostnames defined in the Gateway.PublicGateways (opens new window) configuration map.

Learn more about Kubo configuration for hosting a public gateway:

Known issues

  1. Some browsers and other user agents force lowercase for the authority part of URLs, breaking case-sensitive CIDs before the HTTP gateway has a chance to read them.
  2. DNS label length is limited to 63 characters (RFC 1034 (opens new window))

Due to these limitations, the use of short, case-insensitive CIDv1 in a subdomain context is advised. Base32 is the safe default; the less-popular Base36 can be used for longer ED25519 libp2p keys.

See the next section to learn how to convert an existing CIDv0 to a DNS-safe representation.

# CID conversion for subdomains

If you have content identified by an older CIDv0, there are an automatic and a manual option to safely represent it as CIDv1 for use in subdomains and other case-insensitive contexts.

# Automatic — leverage the gateway in Kubo

Using a subdomain gateway as a drop-in replacement for a path one removes the need for manual CID conversion.

Requests for a content path sent to the gateway domain will return an HTTP 301 redirect to a correct subdomain version, taking care of any necessary encoding conversion if needed:

https://<gateway-host>.tld/ipfs/<cid> -> https://<cidv1>.ipfs.<gateway-host>.tld/

For example, opening the CIDv0 resource at https://dweb.link/ipfs/QmT5NvUtoM5nWFfrQdVrFtvGfKFmG7AHE8P34isapyhCxX/wiki/Mars.html (opens new window) returns a redirect to a CIDv1 representation at https://bafybeicgmdpvw4duutrmdxl4a7gc52sxyuk7nz5gby77afwdteh3jc5bqa.ipfs.dweb.link/wiki/Mars.html (opens new window).

The gateway converts the CID to case-insensitive encoding. The multihash in CIDv1 is the same as in the original CIDv0.

# Manual — use cid.ipfs.io or the command line

The conversion can also be done manually.

To convert a CID to Base32 with no padding (RFC4648 (opens new window)), use cid.ipfs.io (opens new window), or the command line. Below is an example using the command line:

ipfs cid base32 QmbWqxBEKC3P8tqsKc98xmWNzrzDtRLMiMPL8wBuTGsMnR

The output of this is:


PeerIDs can be represented as a CID with libp2p-key multicodec (opens new window). Base36 is suggested as a safer default for longer keys:

ipfs key list -l --ipns-base base36
k51qzi5uqu5dh9ihj4p2v5sl3hxvv27ryx2w0xrsv6jmmqi91t9xp8p9kaipc2 self

ipfs cid format -v 1 -b base36 --codec libp2p-key QmNnooDu7bfjPFoTZYxMNLWUQJyrVwtbZg5gBMjTezGAJN

The gateway provided by Kubo understands the Host header present in HTTP requests and will check if DNSLink exists for a specified domain name (opens new window). If DNSLink is present, the gateway will return content from a path resolved via DNS TXT record. This type of gateway provides full origin isolation (opens new window).

An example is this website, https://docs.ipfs.tech (opens new window).


For a complete DNSLink guide, including tutorials, usage examples, and FAQs, see dnslink.io (opens new window).

# Native URLs

The native address format is the same as a subdomain gateway (opens new window) HTTP URL, but with two differences:

  • The protocol scheme is replaced by the ipfs or ipns namespace
  • The location-based authority component (the gateway host and port) is replaced with a CID





Our main goal here is to reuse existing standards that maximize interoperability with existing user-agents like browsers and CLI tools. If something is not clear, HTTP URL rules apply.

The first element after the double slash is an identifier representing the content root. It is interpreted as an authority component used for origin calculation, which provides necessary isolation between security contexts of different content trees.



Avoid case-sensitive CID

Some user agents will force-lowercase the CID component of URL-like address. To ensure interoperability with existing libraries and software, use case-insensitive CID encoding. Use of CIDv1 in Base32 or Base36 is advised.

# Turning native address to a canonical content path

Every "URL" address can be turned back into a content path with ease:


  • ipfs://{immutable-root}/path/to/resourceA converts to /ipfs/{immutable-root}/path/to/resourceA
  • ipns://{mutable-root}/path/to/resourceB converts to /ipns/{mutable-root}/path/to/resourceB

# Further resources

# Technical specification for implementers

See the IPFS in-web-browsers repository (opens new window).

# Background on address scheme discussions

Discussions around IPFS addressing have been ongoing since @jbenet (opens new window) published the IPFS whitepaper (opens new window), with a number of other approaches being proposed. This long-standing design discussion includes many lengthy GitHub issue threads, but a good summary can be found in this PR (opens new window).

# IPFS Companion

IPFS Companion (opens new window) is a browser extension that simplifies access to IPFS resources.

It provides support for native URLs and will automatically redirect IPFS gateway requests to your local Kubo daemon so that you are not relying on or trusting remote gateways.