You can run a full Bitcoin node behind a residential IP, broadcast a transaction, and have no way to stop your internet service provider, your network operator, or any observer between you and the Bitcoin network from linking that IP address to the transaction you just sent. Until now, if you wanted to hide that link, you had to manually configure Tor routing, install a separate proxy, or trust a third party to do it for you. Bitcoin Core 31.0, released on April 20, 2026, changes that. It adds a single on-off toggle: broadcast all your node's outbound transaction relays exclusively through Tor or I2P. Your IP address never touches the clearnet. Chain analysis firms lose a primary vector for correlation.
This is significant because the link between IP address and transaction timing has been one of the most reliable fingerprints in blockchain surveillance. If you broadcast a transaction at 14:37 UTC and a chain analysis firm sees that same transaction hit a public node 200 milliseconds later from an IP block registered to a major ISP, that is a strong signal. Run millions of these correlations and you start to map the network graph of who owns what. Tor and I2P disrupt that timeline. They add jitter, route through multiple relays, and hide your origin. The Bitcoin network has supported Tor relay since 2012, but running it required deliberate configuration. You had to know it existed, download Tor separately, and manually route your node's traffic. Most node operators never bothered. Bitcoin Core 31.0 makes it default-available and simple enough that a non-technical user can enable it in a GUI or with a single startup flag.
The release also ships a second major change: cluster mempool, a complete architectural redesign of how Bitcoin nodes organize unconfirmed transactions. For years, Bitcoin's mempool has tracked transactions in a flat pool with no internal structure beyond parent-child relationships. This made it hard for nodes to answer a simple question: if I have 1 MB of block space left, which transactions should I include to maximize fee? Cluster mempool organizes all unconfirmed transactions into bounded groups of up to 64 transactions or 101 kilobytes each. Within a cluster, transactions are sorted by their fee rate relative to their descendants. A node can now examine clusters in order and know exactly which transactions to pick. The change improves fee estimation granularity too: the minimum fee rate bucket has been tightened from 1 sat/vB to 0.1 sat/vB, matching the node's default minimum relay fee. This means the fee estimator can now return more precise recommendations for low-fee transactions without having to guess or round up. For casual wallet users, this translates into better on-chain fee predictions and fewer overpaid transactions.
The conditions for both features converging in a single release trace back years. Cluster mempool has been in development across multiple Bitcoin Core maintainers since at least 2022, with code reviews stretching into 2025. The Tor broadcast feature builds on infrastructure that has been in place since v0.12.0 added onion service support in 2014, but integrating it cleanly into the default broadcast stack required careful design work to avoid leaking metadata through other channels like DNS lookups or peer connection patterns. Both landed together because their maturity timelines aligned and because shipping them simultaneously signals a clear priority: the Bitcoin base layer is hardening its privacy story in parallel with performance improvements. Fedimint's v0.11.0 release on April 17 reinforces this: Fedimint gateways can now run behind NAT without firewall configuration, using Iroh for peer discovery. The stack is tightening privacy and self-hosting at every layer.
But the release also removes features that some contracts depended on. The CPFP Carveout, an exception that allowed certain child-pays-for-parent transactions to bypass mempool carveout limits and get processed even when the mempool was full, is now gone. This was a targeted exception designed for Lightning channel closes and other time-sensitive contract operations. Lightning implementations like LND, Core Lightning, and Eclair used this guarantee to ensure that urgent fee bumps would propagate even during mempool congestion. With v31.0, those guarantees no longer exist. Any Lightning node operator running v31.0 must migrate to an alternative fee-bump scheme, likely using replace-by-fee (RBF) or ephemeral anchors. Core Lightning and LND maintainers will need to ship compatibility notices quickly. The -paytxfee and settxfee options, which allowed wallet users to set a static fee rate and then fire off transactions at that rate indefinitely, were also removed after being deprecated in v30.0. These were foot-guns: users could set a rate, forget they set it, and end up overpaying by orders of magnitude during a fee spike, or underpaying and waiting weeks for confirmation. Removing them forces wallet developers to implement per-transaction fee selection, which is safer and gives users more control.
The real read here is that Bitcoin's reference implementation is moving away from simplicity and toward precision. Cluster mempool is not a hack; it is a structural reckoning with the fact that nodes need to solve an optimization problem, not just relay transactions randomly. Tor broadcast is not a bandage; it is an acknowledgment that IP privacy is fundamental to financial privacy and should be baseline, not optional. And the removals matter: contracts and wallets that relied on carveout exceptions or static fee logic now have to adapt. This is what maturity looks like in protocol software. You stop papering over problems and start fixing the foundation. For a user who runs their own node and cares about privacy, v31.0 is the release that makes it easy to do both at once. For a Lightning operator or a smart contract developer, it is a signal: read the migration guide, test on signet, ship your fixes early.
Three things to watch: First, how fast do node-hosting platforms like Umbrel, Start9, and RaspiBlitz ship v31.0 packages. If they move within a week, Tor broadcast adoption could reach tens of thousands of residential nodes within a month. If they take months, adoption will be slower and skewed toward more technical operators. Second, watch for Lightning network advisories. LND and Core Lightning maintainers must publish clear migration paths for CPFP Carveout users, and that work should start now. Any delay risks incompatibility or force-closes during fee spikes. Third, monitor whether the fee estimator changes actually reduce overpayment in wallets. If third-party analysis shows that wallets using Bitcoin Core v31.0's fee estimator consistently undershoot actual fees, that is a signal the change needs refinement. Otherwise, this is the release where Bitcoin's node software stops pretending privacy and precision are optional extras.