// the find
ipfs/kubo
An IPFS implementation in Go
Kubo is the reference Go implementation of IPFS — the peer-to-peer content-addressed filesystem. It's the daemon that most IPFS tooling is built on top of, and it's what you run when someone says 'run an IPFS node'. Target audience is anyone building on IPFS or operating infrastructure that needs to store/retrieve content-addressed data.
Active maintenance with regular releases and a dedicated team (Shipyard) keeping it alive. The HTTP Gateway and trustless retrieval support are genuinely well-specced — you can serve verifiable content to browsers without trusting the gateway operator. Prometheus metrics, structured config, and proper Docker images mean it's not a nightmare to operate. The Bitswap + DHT combination gives you real peer discovery without a central coordinator.
Memory requirements are brutal — 6 GB RAM minimum, and 1 GiB per 20 million pinset items for reproviding. That's not a warning you can ignore; running it on a cheap VPS will cause data to silently disappear from the network. FUSE mounts are still experimental after years. The config surface is enormous and poorly documented for non-obvious interactions — you will spend time debugging why your node isn't announcing content correctly. IPFS itself has struggled with adoption; you're betting on a network effect that hasn't fully materialized.